Characteristics and Perceptions of Cost-Share Funding for Emerald Ash Borer Mitigation in Virginia Urban Areas

Characteristics and Perceptions of Cost-share Funding for Emerald Ash Borer Mitigation in Virginia Urban Areas

Peter W. Stewart

Thesis submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of

Master of Science In Forestry

P. Eric Wiseman, Chair

John F. Munsell

Scott M. Salom

May 23, 2019

Blacksburg, Virginia

Keywords: invasive species management, forest health, incentive program

Characteristics and Perceptions of Cost-share Funding for Emerald Ash Borer Mitigation in Virginia Urban Areas

Peter W. Stewart

ABSTRACT (ACADEMIC)

Since most invasive forest pests first establish in urban areas, detection and containment of these pests within cities is important to the health of all forests. While the emerald ash borer

(EAB) (Agrilus planipennis Fairmaire) has proved difficult to contain, efforts continue to mitigate the impacts of its spread. As part of those efforts, the Virginia Department of Forestry

(VDOF) initiated its Emerald Ash Borer Treatment Program (EABTP) in 2018, providing financial incentives for insecticidal protection of ash trees. To better understand the role of incentives in promoting urban forest health, I conducted a study of properties, households, and practitioners involved in the program’s first year.

To examine where EABTP funding helped pay for tree protection, I conducted tree inventories on 16 urban participant properties. Concurrently with tree inventory work, I conducted web and mail surveys to examine homeowner engagement in preservation of threatened trees. Finally, to investigate the role of forest practitioners involved in program implementation, I conducted web surveys of VDOF foresters and Virginia arborists. Results demonstrated that on urban participant properties—typically large and wooded—white ash

(Fraxinus americana) was the dominant species. Results from homeowner surveys demonstrated broad support for personal investment in tree preservation, and the significance of attitudinal predictors towards those intentions. Results from practitioner surveys demonstrated substantial, though not unanimous, support for the program as a benefit both to clients and forests.

Implications of these findings are discussed in the context of future urban forest health initiatives.

Characteristics and Perceptions of Cost-share Funding for Emerald Ash Borer Mitigation in Virginia Urban Areas

Peter W. Stewart

ABSTRACT (PUBLIC)

Because most non-native forest pests arrive in cities before spreading to rural areas, detecting and containing these pests within urban forests is important to all forested areas. One non-native pest, the emerald ash borer (EAB), has proved difficult to contain, but there are ongoing efforts to limit the damage it causes as it spreads. As part of those efforts, the Virginia

Department of Forestry (VDOF) began its Emerald Ash Borer Treatment Program (EABTP) in

2018, which offered partial reimbursement for the cost of protecting ash trees with insecticide.

To better understand how reimbursement payments might help promote the health of urban trees,

I studied the properties, households, and practitioners involved in first year of the program.

To examine where EABTP funding helped pay for tree protection, I conducted inventories of all trees on 16 participating properties in urban areas. At the same time, I conducted web and mail surveys to examine how homeowners thought about urban tree preservation. Finally, I conducted web surveys of VDOF foresters and Virginia arborists, to investigate roles of these practitioners in implementing the program. Results indicated that on urban participant properties, which were typically large and wooded, white ash was the dominant species. Results from homeowner surveys demonstrated broad support for personal investment in tree preservation, and the significance of attitudes in predicting that support. Results from practitioner surveys demonstrated substantial, though not unanimous, support for the program as a benefit both to clients and forests. These findings are discussed in the context of future urban forest health programs.

ACKNOWLEDGEMENTS

Many thanks to Dr. Eric Wiseman for guidance and encouragement throughout my entire graduate program. Thanks also to committee members Dr. Scott Salom and Dr. John Munsell for providing important feedback and support throughout the process of study design, survey research, and analysis. A big thanks to Lori Chamberlin, Meredith Bean, and all at Virginia

Department of Forestry’s Forest Health and Urban & Community Forestry divisions for collaboration and distribution of survey requests. I’m extremely grateful for the help of John

Peterson at Virginia Tech in conducting tree inventories on properties around the state, in mostly frigid conditions. Thanks also to Sarah Gugercin, Tiffany Brown, and David Reep at Virginia

Tech for the time and expertise they contributed towards survey layout and mailing.

Additionally, I’m grateful for the help of many of my graduate student colleagues with technical aspects of geographic and survey data analysis. I also greatly appreciate the contributions of time and personal input from all homeowners, arborists, and foresters who participated in surveys or property inventories. Finally, I’m grateful for the constant support my wife Laura has provided during this project.

TABLE OF CONTENTS

LIST OF FIGURES ...... x

LIST OF TABLES ...... xi

CHAPTER 1 – INTRODUCTION ...... 1 1.1 Research Background ...... 1 1.2 Research Objectives ...... 2

CHAPTER 2 – LITERATURE REVIEW ...... 4 2.1 Biological Invasions in North America ...... 4 2.2 Historical Impacts of Invasive Forest Pests and Pathogens ...... 5 2.3 Arrival, Spread, and Impacts of the Emerald Ash Borer ...... 8 2.4 Biology and Control of EAB ...... 12 2.4.1 Strategies for Municipal EAB Management ...... 14 2.4.2 State-level EAB Management Cost-share Programs ...... 15 2.5 Historical and Current Context of Incentive Programs ...... 16 2.6 Analysis of Incentive Programs and Landowner Participation ...... 18 2.7 Summary of Literature ...... 22

CHAPTER 3 – SITE CHARACTERISTICS OF URBAN EABTP PARTICIPANT PROPERTIES ...... 24 3.1 Introduction ...... 24 3.2 Methods ...... 25 3.2.1 Definition of Target Population ...... 25 3.2.2 Selection of Sites...... 25 3.2.3 Data Collection: Participant Properties ...... 26 3.2.4 Data Processing: i-Tree Eco...... 25 3.2.5 Data Compilation: Urban Forest Assessments ...... 259 3.2.6 Data Analysis ...... 30 3.3 Results ...... 31 3.3.1 Data Validation ...... 31 3.3.2 Cluster Analysis of Participant Properties ...... 32 3.3.3 Comparison of Species Composition Among Sites ...... 35

3.3.4 Comparison of Species Composition Between Sites and Surrounding Urban Forests...... 36 3.3.5 Linear Models of Species Composition ...... 39 3.4 Discussion ...... 41

CHAPTER 4 – HOMEOWNER ENGAGEMENT IN LANDSCAPE TREE PRESERVATION ...... 43 4.1 Introduction ...... 43 4.2 Methods ...... 45 4.2.1 Construction of Survey Sampling Frames ...... 45 4.2.2 Construction and Pre-testing of Survey Instruments ...... 49 4.2.3 Data Collection ...... 50 4.2.4 Data Analysis ...... 53 4.3 Results ...... 58 4.3.1 Summary of Response Data ...... 58 4.3.2 Tests for Nonresponse Bias ...... 61 4.3.3 Analysis of Response Mode...... 63 4.3.4 Factor Analysis of Attitudes Towards Urban Trees ...... 64 4.3.5 Cluster Analysis of Ranked Motivations for Tree Preservation ...... 64 4.3.6 Linear Models of Behavioral Intention ...... 66 4.4 Discussion ...... 70

CHAPTER 5 – PERCEPTIONS OF COST-SHARE PARTICIPATION AMONG FOREST PRACTITIONERS ...... 72 5.1 Introduction ...... 72 5.2 Methods ...... 74 5.2.1 Construction of Survey Sampling Frames ...... 74 5.2.2 Construction of Survey Instruments ...... 75 5.2.3 Data Collection ...... 76 5.2.4 Data Analysis ...... 77 5.3 Results ...... 81 5.3.1 Summary of Response Data ...... 81 5.3.2 Tests for Nonresponse Bias ...... 84 5.3.3 Factor Analysis of Attitudes Towards Urban Trees ...... 85 5.3.4 Cluster Analyses: Motivations, Recommendations, and Pest Risk ...... 86

5.3.5 Chi-square Analysis of Pest Risk Perceptions ...... 87 5.3.6 Linear Models of Program Participation ...... 88 5.3.7 Qualitative Analysis of Written Responses ...... 90 5.4 Discussion ...... 91

CHAPTER 6 – THESIS CONCLUSION ...... 93 6.1 Summary of Findings ...... 93 6.1.1 Urban Participant Properties ...... 93 6.1.2 Homeowner Engagement in Landscape Tree Preservation ...... 94 6.1.3 Practitioner Perceptions of Cost-Share Participation ...... 95 6.2 Implications ...... 96 6.2.1 Urban Participant Properties ...... 96 6.2.2 Homeowner Engagement in Landscape Tree Preservation ...... 100 6.2.3 Practitioner Perceptions of Cost-share Participation ...... 103 6.3 Study Limitations ...... 105 6.4 Future Research ...... 108

REFERENCES ...... 111

APPENDIX A - Supplementary Tables for Chapter 3 ...... 135

APPENDIX B - Supplementary Tables for Chapter 4 ...... 140

APPENDIX C - Supplementary Tables for Chapter 5 ...... 151

APPENDIX D - 2018 EABTP Application Form ...... 158

APPENDIX E - Tree Inventory Data Collection Form ...... 160

APPENDIX F - Survey Recruitment Materials ...... 161

APPENDIX G - Survey Instruments ...... 165

APPENDIX H – Western IRB Exemption Letter ...... 193

vii

LIST OF FIGURES

Figure 1. Map of the range of ash species in the U.S. and the extent of the EAB infestation, as of April 2019 ...... 10

Figure 2. Map of relative abundance of ash species, by county, in Virginia before the arrival of EAB, drawn from 1998-2002 FIA data ...... 11

Figure 3. Map showing the spread of EAB across Virginia, with Census Urban Areas displayed and rough locations of 2018 EABTP participants ...... 11

Figure 4. Dendrogram of urban participant properties from a hierarchical cluster analysis of parcel tree cover, parcel size, and age of property development ...... 34

Figure 5. Relative abundance of species across 16 EABTP participant properties. Top ten species by relative abundance are shown ...... 38

Figure 6. Relative structural value of species across 16 EABTP participant properties. Top ten species by relative structural value are shown ...... 38

Figure 7. Relative abundance of species across 3 Virginia municipalities. Data were drawn from 2010-2011 Urban Forest Assessments of Falls Church, Charlottesville, and Roanoke .... 38

Figure 8. Relative abundance of species across 3 Virginia municipalities. Data were drawn from 2010-2011 Urban Forest Assessments of Falls Church, Charlottesville, and Roanoke .... 38

Figure 9. Map of EAB infestation strata, Census Urban Areas, and study areas for the general household survey ...... 48

Figure 10. Map of Virginia displaying counties by EAB infestation strata, and rough locations of forester and arborist survey respondents ...... 84

viii

LIST OF TABLES

Table 1. Characteristics of all EABTP urban participant properties ...... 32

Table 2. Property characteristics by cluster membership for 16 inventoried urban participant properties...... 35

Table 3. ANOVA comparison of top ten species by relative abundance and relative structural value across 16 inventoried urban participant properties ...... 37

Table 4. Results from exact sign test of species composition between 13 urban participant properties and surrounding urban forests ...... 37

Table 5 Summary of modeled ash importance and structural value across 16 urban participant properties in the 2018 EABTP ...... 40

Table 6. Total population, sampling frame size, and number of survey recipients, by study area for general household survey ...... 48

Table 7. Summary of selected survey responses and external data for program participants and general households ...... 62

Table 8. Binomial logistic regression results of mail vs. web responses to the general household survey, using demographic predictors ...... 63

Table 9. Cluster analysis of ranked statements about motivations for tree preservation ...... 65

Table 10. Program participants: Ranking of candidate models for the dependent variable Tree preservation intention ...... 67

Table 11. Program participants: Summary of top-ranking model Attitudes for the dependent variable Tree preservation intention ...... 67

Table 12. General households: Ranking of candidate models for the dependent variable Tree preservation intention ...... 68

Table 13. General households: Summary of top-ranking model Attitudes + Property Characteristics for the dependent variable Likelihood of regularly treating threatened shade trees ...... 68

Table 14. Summary of selected survey responses and external data for foresters and arborists .. 83

Table 15. Aggregated forester and arborist response data: Cluster analysis of forest pests ranked by level of perceived threat to Virginia's forests...... 87

Table 16. Chi-square analyses of independence between practitioner pest threat clusters and professional characteristics ...... 88

Table 17. Arborist response data: Candidate models for dependent variable Interest in future program participation ...... 89

Table 18. Arborist response data: Summary of model Professional characteristics for the dependent variable Interest in future program participation ...... 90

Table 19. Arborist response data: Summary of model Attitudes for the dependent variable Interest in future program participation ...... 90

CHAPTER 1 – INTRODUCTION

1.1 Research Background

Pests and pathogens present an immense threat to North American forests, having caused greater economic and ecological impact than wildfire, drought, or severe weather (Liebhold

1995). Invasions of non-native forest pests typically orginate in cities (Paap et al. 2017) for several reasons: cities function as transportation nodes for freight which may tranport invasive species; cities also contain both a high proportion of non-native tree species which may host non- native pests, and a high proportion of stressed trees susceptible to attack (Pautasso et al. 2015).

Damages to urban forests from a single invasive pest may reach into the tens of billions of dollars (Kovacs et al. 2010); economic and ecological damage can compound when infestations spread into surrounding rural forests (Lovett et al. 2016). For these reasons, forestry officials, commercial arborists, and urban residents all have important roles in the detection and containment of invasive forest pests.

More than 450 invasive forest pests and 16 pathogens known to cause damage have been documented in North America since 1860 (Aukema et al. 2010). Since its arrival in the late

1990s, emerald ash borer (EAB) (Agrilus planipennis Fairmaire) has become the most economically destructive of all of these (McCullough and Mercader 2012). It important to note, however, that a comparison of damage caused by invasive forest pests does not include damage caused by either the southern pine beetle (Dendroctonus frontalis Zimmerman) or the mountain pine beetle (Dendroctonus ponderosae Hopkins), both native to North America. In addition to creating widespread economic and ecological damage, EAB causes such high mortality rates that entire native ash species (Fraxinus spp., Family: Oleaceae) are threatened with extirpation

(Lovett et al. 2016). For many areas of the eastern U.S. where the ash borer has been present for

several years, there may be few ash trees left to preserve. Yet for other communities where EAB has recently been detected or is projected to arrive, insecticidal treatment of ash trees may yield cost savings compared to tree removal, retention of benefits provided by mature urban trees

(Sadof et al. 2017), and the continued genetic viability of seed-producing trees.

In 2018, Virginia Department of Forestry (VDOF) began offering cost-share payments for the insecticidal treatment of ash trees through its Emerald Ash Borer Treatment Program

(EABTP). The program aimed to incentivize treatment of valuable ash trees on public and private lands among those for whom the initial expense of treatment might be a barrier (VDOF

2018a). Additionally, since no treatments provide indefinite protection, this program also aimed to ‘jumpstart’ continued, independent treatment among participants (Chamberlin 2018a). Unlike similar programs administered in neighboring states, Virginia’s EABTP accepted applications from individual homeowners, in addition to organizations and municipalities. This program completed its first funding period in August 2018, with 107 total participants, of whom 90 were homeowners. The remaining 17 participants were comprised of 6 nonprofit or neighborhood organizations, 5 municipalities, 3 businesses, and 3 educational institutions.

1.2 Research objectives

Because of its unique participant pool, the EABTP provided an important opportunity for studying the influence of financial incentives on the management of urban forest pests, particularly in residential areas. Through this research, my intent was to provide insight into outcomes of the 2018 EABTP and the potential application of incentive programs for management of other urban forest pests. There were two facets to my research: (1) conducting tree inventories of urban participant properties, and (2) conducting survey research with the two

primary EABTP stakeholder groups of homeowners and forest practitioners. Primary stakeholder groups were in turn each comprised of two discrete sampling frames: among homeowners, I contacted program participants, and a sample of general households. Amont practitioners, I contacted VDOF county foresters and arborists within Virginia. My research objectives and questions were:

1) To characterize urban EABTP participant properties.

• What types of properties commonly received funding?

• What contributions did ash trees make to the forest composition and appraised landscape

value of participant properties?

• Are property characteristics associated with species composition of landscape trees?

2) To examine urban homeowner engagement in preservation of threatened urban trees.

• Are attitudes towards tree preservation different between program participants and

general households?

• What characteristics are most strongly associated with homeowners’ willingness to pay

for preservation of threatened trees?

• What characteristics are most strongly associated with interest in cost-share participation?

3) To examine forest practitioners’ engagement in preservation of threatened urban trees.

• Are specific professional characteristics associated with differing perceptions of threats

posed by forest pests?

• What characteristics are most strongly associated with interest in cost-share participation?

• What are the most commonly-cited reasons for interest or lack of interest in the EABTP?

CHAPTER 2 – LITERATURE REVIEW

2.1 Biological Invasions in North America

Non-native species are those which humans have transported from one region to another; precisely defining which species are not native to a region can be complex (Richardson and

Pyšek 2004). In the western hemisphere, native species are typically defined as those present in the Americas prior to European settlement (NRCS 2019). This date coincides with the beginning of an era of transoceanic trade—an unprecedented movement of people, goods, and biota. Today, despite regulation concerning inadvertent transport of biotic material across oceans or state lines, there are well over 50,000 non-native species in the U.S. alone (Pimentel et al. 2005). Not all non-native species are necessarily invasive; this label is applied to those that cause unacceptable economic damage or threaten local biodiversity or ecosytems (Schmiedel et al. 2016).

Biological invasions are often conceptualized as occurring in stages; with each succeeding stage, fewer introduced species survive. Four stages of invasion commonly described are long distance transport, colonization, establishment, and geographic spread (Theoharides and

Dukes 2007). A rule of thumb proposed by Williamson (1996) states that only 10% of species survive each successive transition, such that for every 1000 species in transport, only 1 will survive long enough to spread across a region. Since survival of non-native species in unfamiliar territory is somewhat counterintuitive, several hypotheses attempt to account for their apparent success in new bioregions. Two of the most well-known hypotheses are enemy release, which credits an invader’s success to a lack of co-evolved enemies, and its converse, novel weapon, in which prey or host organisms of the invaded environment possess no evolved defenses against the invader’s attack mechanisms (Alpert 2006). Through these pathways or others, the number of non-native species in North America continues to grow—among forest pests only, at least two

new species become established annually (Aukema et al. 2010). Of invasive forest pests currently present in the U.S., wood- and phloem-boring insects (primarily Coleopterans) account for few species overall, but account for a large share of economic damage, and are becoming established at an increasing rate (Aukema et al. 2011). For most forest pests, the first three stages of invasion take place between and within cities; spread to forested rural areas is likely only after a pest or pathogen has established itself on urban host trees (Paap et al. 2017).

2.2 Historical Impacts of Invasive Forest Pests and Pathogens

While emerald ash borer (EAB) is judged to be the most destructive of any invasive forest pest in North America (McCullough and Mercader 2012), it is not unique in causing widespread economic and ecological damage (Aukema et al. 2011). Additionally, quantitative comparisons of damage to forests caused by invasive species often omit consideration of invasive forest pathogens, whose impacts in North America have been equally, if not more widespread than those of insects (Loo 2009). A discussion of EAB’s 21st century impact is better understood, then, in the context of invasive forest pests and pathogens in North America over a much longer period. For that purpose, included here are descriptions of four forest invaders (two pests and two pathogens), each of which likely established in an urban area and subsequently caused widespread damage to North American forests.

One of the earliest and most well-known examples of an invasive forest pest infestation in North America is that of the gypsy moth (Lymantria dispar L.). This lepidopteran foliar- feeder, native to Europe, was brought to Medford, MA in the late 1860s by French amateur scientist Etienne Leopold Trouvelot, and soon escaped through the window of his lab (Liebhold et al. 1989). Within 10 years, moths were heavily defoliating the neighborhood, and by 1890 the

State of Massachusetts was appropriating money for its control, followed by the Federal

Government in 1906 (McFadden 1991). The 20th century saw repeated outbreaks extending over millions of square miles of forest, and several concerted management campaigns, including the current ‘Slow the Spread’ campaign, initiated in 1995 (McManus and Csóka 2007). Through this campaign the USDA Forest Service and partner agencies have been largely successful in containing the gypsy moth to the northeastern quadrant of the country through coordinated trapping and aerial spraying of pheremonal disruptants (Sharov et al. 2002).

A second long-established and devastating North American forest invader is chestnut blight, caused by Cryphonectria parasitica (Murrill) M.E. Barr, a fungal pathogen likely transported from its eastern Asian range on nursery stock in the early 20th century (Griffin 2000).

Within 50 years of its 1904 identification in the Bronx, NY, this pathogen spread through the entire range of the American chestnut (Castanea dentata (Marshall) Borkh.), virtually wiping out the species as an overstory tree (Rigling and Prospero 2018). Stump sprouts from trees killed decades prior still grow today, although they are inevitably killed off by the pathogen before reaching sexual maturity (Paillet 2002). Restoration of the American chestnut, once dominant in eastern North American forests, has been for decades the ongoing work of researchers dedicated to developing resistant hybrids and cultivars (Clark et al. 2019).

Like chestnut blight, Dutch elm disease is caused by rapidly spreading forest pathogens

(Ophiostoma ulmi (Buisman) Nannf. and O. novo-ulmi Brasier), although their region of origin is unclear (Brasier 1990). Diseased trees were first recorded in Cleveland, OH in 1930, thought to be infected from veneer logs imported from Europe (Schlarbaum et al. 1997). Attacking primarily American elm (Ulmus americana L.), these pathogens also spread quickly, vectored by multiple bark beetle species, and encompassed the elm’s native range within 50 years. While

some elms in forest settings displayed resistance, urban cultivars succumbed rapidly to the disease, leading to a dramatic loss of tree cover in many midwestern and northeastern cities, where American elms had been heavily planted as street trees (Campanella 2003). White and green ash (Fraxinus americana L. and F. pennsylvanica Marshall) were often planted as replacement trees, because of tolerance to urban conditions and attractive form (MacFarlane and

Meyer 2005).

In 1996, the Asian longhorned beetle (ALB) (Anoplophora glabripennis Motschulsky) was first identified in Brooklyn, New York, noticed for the dime-sized exit holes left on tree trunks (Haack et al. 1996). A separate infestation was detected soon after in Chicago, followed by others in urban areas of New Jersey, Massachusetts, and Ontario (Dodds and Orwig 2011).

Because this wood-boring cerambycid was observed to kill a wide ranges of host trees, including common urban genera such as Acer, Aesculus, and Salix, eradication campaigns quickly began in all affected locations (Haack 2006). These campaigns, funded by USDA-APHIS, consisted of outreach, detection efforts, destruction or treatment of potentially infested trees within a buffer zone, and an enforcement of quarantine on potentially infested products (Smith et al. 2001).

While new infestations have occurred in the northeastern U.S. and as far away as Sacramento,

CA, several of the earliest infestations have been declared eradicated by APHIS (American

Nurseryman 2013). The relative level of success achieved in containing these infestations stands in contrast to the unchecked spread of the emerald ash borer. However, the potential for widespread damage remains—because of the number of ALB host species, a widespread infestation in urban forests alone could cause 30% tree mortality and a corresponding loss of

$669 billion in value (Nowak et al. 2001).

2.3 Arrival, Spread and Impacts of the Emerald Ash Borer

Emerald ash borer, a phloem-feeding buprestid beetle native to eastern Asia, is several times smaller than ALB in its adult stage and more difficult to detect. While EAB was first identified in the Detroit metropolitan area in 2002, dendrochronological analysis shows it may have been present as early as 1997 (Siegert et al. 2014). The invasion’s likely epicenter was an industrial area of Wayne County, MI, close to interstate highways, airports, and shipping terminals. Within fifteen years of arrival, the beetle had created more economic damage to U.S. forests than any other invasive pest in the nation’s history (McCullough and Mercader 2012). By

2019, 22 years after its estimated arrival, EAB populations had spread to cover much of the range of eastern North American ash (Fraxinus spp., Family: Oleaceae), killing millions of trees across 35 U.S. states and 5 Canadian provinces (EAB Information Network 2019). Significant ecological effects are also expected among bird, mammal, and insect species which rely on ash species for food or habitat (Liu 2018). Figure 1 displays the 2019 North American extent of the

EAB infestation and the historical range of ash species.

In its native east Asian range, emerald ash borer has only been observed attacking declining ash trees. In North America, however, it aggressively infests healthy trees, including all 16 ash species native to the continent (Cappaert et al. 2005). Resulting ash mortality rates are typically over 99% in heavily infested areas (Knight et al. 2013). Successive attempts to eradicate or contain infestations have been unsuccessful, primarily because of the difficulties of detecting low-density populations and enforcing quarantine regulations (Liu 2018).

Economic impacts of EAB damage are comparatively highest in urban areas, where costs to local governments and homeowners are many times greater than timber value losses in rural areas (Aukema et al. 2011). Urban costs include insecticidal ash treatment, removal and

replacement, in addition to associated property value loss. Kovacs et al. (2010) projected a $10.7 billion loss due to EAB in urban areas alone, for a 25-state area over the period 2010-2019. This figure did not include losses to property value and significantly underestimated the EAB’s rate of spread over that period, indicating true costs may be higher. High costs to urban areas are a function of the prevalence of white and green ash as landscape and street trees. Ironically, ash trees were most common in northeastern and midwestern cities where they had been used to replace American elm trees lost to Dutch elm disease (MacFarlane and Meyer 2005).

Until the arrival of EAB, native ash species made up 1.5% of Virginia’s forested lands, ranging from over 6% relative abundance in Northern Piedmont counties to essentially 0% in some coastal locations, shown in Figure 2 (FIA 2019). These data are drawn from the five-year

FIA sampling cycle 1998-2002, and represent relative abundance of ash species in rural forests, prior to the arrival of the ash borer in the state. Significantly, relative abundance of native ash species among street tree populations mirrored that of forested lands. In one study, researchers conducted street tree inventories and compiled street tree inventory data for a total of fourteen

Virginia municipalities, spanning the years 2003-2010. Among these cities, distributed across

Virginia bioregions, native ash relative abundance ranged from 0.1% to 5.8%, and averaged

2.0% across studied cities (Wright 2011).

EAB was introduced to Virginia via nursery stock within a year of its initial 2002

Michigan detection. Virginia forestry officials initiated a rapid eradication campaign at the infested site, destroying all ash trees within a half-mile radius and instituting a meticulous detection program (Asaro 2007). These efforts temporarily provided an example of successful

EAB eradication, until the summer of 2008 when the beetle was again detected in multiple

Fairfax County sites. Eradication was abandoned in favor of a strategy of containment within a

five-county quarantine zone (Asaro 2008). Quarantine of northern Virginia counties in turn was abandoned by 2012, at which point widespread detections in Virginia led USDA-APHIS to designate the entire state within the federal EAB quarantine zone (Asaro 2013). By 2019, of 130

Virginia counties and independent cities, only 37 remained without a confirmed EAB detection

(Chamberlin 2019). Figure 3 displays the year of initial EAB detection, by Virginia county.

Figure 1. Map of the range of ash species in the U.S. and the extent of the EAB infestation, as of April 2019.

Figure 2. Map of relative abundance of ash species, by county, in Virginia before the arrival of EAB. Data were drawn from 1998-2002 FIA data.

Figure 3. Map showing the spread of EAB across Virginia, with Census Urban Areas displayed, and rough locations of 2018 EABTP participants.

2.4 Biology and Control of EAB

In Virginia, EAB typically undergoes complete metamorphosis in one year, with adult emergence taking place in early May (Chamberlin 2018b). During the month following emergence, adult beetles feed on ash foliage, mate, and lay eggs in bark crevices (Liu 2018).

Females may lay up to 80 eggs, and show a preference for stressed trees as oviposition sites

(Jennings et al. 2014). Neonate larvae hatch in mid-summer, bore inwards through the bark, and begin to feed in phloem, cambium and outer sapwood, carving characteristically serpentine galleries (Cappaert et al. 2005). Larvae complete four instars while feeding from July until

November, followed by an overwintering stage. Pupation lasts for approximately one month in spring, after which the new adults tunnel out though the bark, leaving a distinctive D-shaped exit hole (Van Driesche and Reardon 2015).

Damage caused to ash trees by foliar feeding of adult beetles is minimal compared to the effects of larval feeding in vascular tissue. As the density of larval galleries increases in a tree, translocation of nutrients and water is greatly reduced—resulting in visible crown thinning and dieback (Herms and McCullough 2014). Within five years of on-site EAB detection, a stand of ash trees is likely to experience a 99% mortality rate (Knight et al. 2013). Declining trees typically display shoot and branch dieback progressing downwards through the crown, and often produce heavy epicormic growth on the trunk and lower crown (Van Driesche and Reardon

2015).

Coordinated regional campaigns to contain EAB using chemical and cultural methods, such as the efforts described in Virginia, have repeatedly proven unsuccessful. Much EAB- related research today is instead directed toward identifying resistance and regeneration capacities of ash species. Studies conducted on site characteristics of ‘lingering ash’ pockets in

the Midwest have noted the higher survival rates of ash trees at a distance from other ash (e.g.,

Knight et al. 2011, Kappler et al. 2018). Others studies have investigated successional dynamics of germination and ‘orphaned’ ash seedlings in EAB-infested forest stands, concluding that silvicultural management is essential to ash survival (Klooster et al. 2018, Granger, Zobel, and

Buckley 2017).

For years, researchers have also investigated the potential for biocontrol of EAB through introduced insect natural enemies. While several native North American bird, insect, and fungi species feed on EAB, none have been observed to significantly suppress its population (Bauer et al. 2015). In its native east Asian range, however, natural enemies can reduce EAB population density by as much as 74% (Liu et al. 2007). Following years of quarantined trials, USDA-

APHIS has approved the U.S. release of four Asian hymenopteran parasitoids (USDA-APHIS

2017), all of which parasitize EAB eggs or larvae at high rates in experimental settings (Jennings et al. 2016). Researchers theorize that biocontrol agents in the future may protect surviving and regenerating ash by reducing the severity of EAB outbreaks (Duan et al. 2018).

While control of EAB populations is not yet feasible, an infestation’s impact may be mitigated on a municipal scale through planned use of tree removal and insecticidal treatment.

Effective protection of individual ash trees can be achieved through the application of systemic insecticides to a tree’s trunk or root system, disrupting larval development in phloem tissue

(Smitley, Doccola, and Cox 2010). In particular, trunk injections of emamectin benzoate can reduce EAB larval densities by 99% and provide protection against infestation for three years

(McCullough et al. 2011). However, high cost of treatment creates a significant barrier to tree treatment both for municipalities and property owners. The commercial cost of emamectin injection can be as high as $15 per inch of trunk diameter (about $6 per centimeter) (Chamberlin

2018b). At this rate, treatment of an 18” DBH tree (46 cm) would cost $288—if extended to the ash population of an entire municipality, costs can quickly reach six or seven figures.

2.4.1 Strategies for Municipal EAB Management

Urban forestry officials facing a recently detected or approaching EAB infestation must weigh short- and long-term costs of multiple management strategies, and importantly, the benefits accrued through these actions. If no action is taken, close to 100% mortality of a municipality’s ash trees can be expected within eight years of initial EAB detection (Sadof et al.

2011). Many municipalities’ urban forestry budgets have been overwhelmed by the volume of tree removal necessitated by EAB, which peaks 5-7 years following initial local detection of the borer (Sadof et al. 2017). With planning, a city’s urban forestry department can distribute the concentrated financial burden of tree removal over many years, and at the same time sustain urban forest value through insecticidal treatment of strategically selected trees.

To help municipalities plan for EAB infestation, researchers have modeled urban forest growth under a range of management scenarios, predicting total costs and resulting net benefits in retained urban forest value. As assessed by benefit-to-cost ratios, model predictions strongly support the use of insecticidal treatment for all or most ash trees, as opposed to options of preemptive tree removal, or removal and replacement (Vannatta, Hauer, and Schuettpelz 2012,

Sadof et al. 2017). Planning may also include outreach and coordination with private property owners, whose properties typically represent 90% of the urban forest (Moll 1993). Spatially- explicit models incorporating privately-owned trees in municipal EAB planning indicated higher overall tree survival rates (McCullough and Mercader 2012) and further improved benefit-to-cost ratios (Kovacs et al. 2014).

2.4.2 State-level EAB Management Cost-Share Programs

Since 2012, forestry and natural resource agencies in several eastern states have developed programs to promote EAB management planning among municipalities and residents.

These programs grew out of research conducted during earlier stages of EAB infestation in the

Midwest (Liu 2018). Programs developed by Pennsylvania, Maryland, Virginia, and North

Carolina offer a combination of training and funding to municipalities who develop detailed

EAB management plans. In all cases, program guidelines advocate for the use of emamectin benzoate, if insecticides are to be a management component. Eligibility guidelines vary, but generally allow only municipalities and other public agencies to apply for assistance. Among these programs, Virginia’s Emerald Ash Borer Treatment Program is unique in accepting applications from private property owners.

The earliest of these programs, Pennsylvania’s Department of Conservation and Natural

Resources’ (PDCNR) Community Ash Management Plan (CAMP) program began accepting applicants in 2012. Participating municipalities or agencies receive training in i-Tree inventory software and EAB biology, and up to $5000 in funding to implement an ash management plan.

Participants must choose from one of four strategies: (1) no action, (2) selective management,

(3) preemptive management, and (4) aggressive management (Liu 2013). In 2017, Maryland’s

Department of Natural Resources-Forest Service (DNR-FS) initiated its Emerald Ash Borer

Treatment Program, a 50/50 cost-share program for ash tree treatment by municipalities, state agencies, and private conservation easements. Like Pennsylvania’s program, applicants must develop a written Community Response Plan, an accompanying tree inventory, and demonstrate need based on public value of trees or potential hazards (MD DNR-FS 2017). North Carolina

Forest Service’s Ash Protection Program, begun in 2018, allows only municipalities to apply for assistance with ash tree treatment. A single municipality may receive up to $5000 for tree

treatment, at the reimbursement rate of $12 per inch diameter (NCFS 2018a). Program guidelines do allow, however, for funds to pay for treatment of trees on private land at local officials’ discretion (NCFS 2018b).

In contrast to eligibility requirements of other programs, Virginia Department of

Forestry’s (VDOF) Emerald Ash Borer Treatment Program (EABTP) funds ash tree treatment among municipalities, private organizations, and individual property owners. Initiated in 2018, this cost-share program offered 50% reimbursement of tree treatment cost, capped at $1250 for individuals and $5000 for organizations or municipalities. Funding priority was designated for applicants from counties with no detected EAB presence, followed by those from counties where first detection occurred post-2014 (VDOF 2018c). EABTP’s application process was designed to require minimal paperwork—application materials include a one-page form signed by a VDOF forester and a bid for tree treatment from a licensed contractor (VDOF 2018c). In its initial year,

EABTP program participants included 90 homeowners, 5 municipalities, and 12 educational, business, nonprofit, or other organizations (Bean 2019).

2.5 Historical and Current Context of Incentive Programs

Cost-share programs such as Virginia’s EABTP have been developed within the broader context of financial incentive programs, administered for decades by state and federal natural resource and conservation agencies. Depending on context, these programs may be described as conservation incentive programs, voluntary incentive programs, financial incentives, or simply cost-share. I will use cost-share and incentive program interchangeably, to refer generically to all types of incentive programs. Programs range in purpose from conservation objectives such as habitat protection to economic concerns like timber supply. What programs share is the objective

to correct a perceived market failure in which private landowners do not have sufficient incentive to produce external benefits from their land, such as food, fiber, or ecosystem services (Caldas et al. 2016). Ecological benefits in particular, may also be described as public goods—those from which all may benefit without diminishing another’s enjoyment (Goldman et al. 2007). Whether intended to stimulate tree planting, soil conservation, or tree preservation, incentive programs aim to compensate private property owners for investment in publicly-beneficial work the landowners either cannot afford or from which they would not realize substantial personal benefit.

Within U.S. forestry, an educational Extension Service was first authorized through the

Clarke-McNary Act of 1924, ten years after the establishment of the Agricultural Extension

(Boyd et al. 1988). Through the Norris-Doxey Cooperative Farm Forestry Act of 1937, direct forestry assistance was first offered in the form of subsidized planting stock and forestry demonstrations (Barden et al. 1996). Program objectives shifted over time from a focus on timber production to include multiple forest management objectives, such as those of the

Forestry Stewardship Program, authorized through the Cooperative Forestry Assistance Act of

1978 (Kilgore et al. 2007). Over the same period, state forestry agencies also began administering incentive programs, such as Virginia’s Reforestation of Timberlands, which has been operating continuously since 1970 (Cumbia 2018).

Other forestry programs have incentivized proactive responses to forest pest outbreaks.

The Southern Pine Beetle Prevention and Restoration Program (SPBRP), a USFS incentive program, has promoted thinning, burning and re-planting across all southern states in response to repeated beetle outbreaks (Nowak et al. 2008). VDOF disburses federal SPBRP funding to

Virginia landowners through its own Pine Bark Beetle Prevention Program (Watson et al. 2013).

Longstanding incentive programs operate within other land management contexts, notably those administered by the USDA Natural Resources Conservation Service (NRCS).

Since 1985, the NRCS’s Conservation Reserve Program has paid farmers to take ecologically sensitive lands out of production, spending $1 billion annually for the program’s first 15 years

(Feng et al. 2005). NRCS programs with similar habitat protection objectives have included the

Wetlands Reserve Program, Grassland Reserve Program, and Working Lands for Wildlife

(Ciuzio et al. 2013).

Despite the history and prevalence of incentive programs within forestry and natural resource management, there is little evidence of their use in urban settings, prior to state-level

EAB management programs. One of the few predecessors to Virginia’s EABTP is Minnesota

Department of Natural Resource’s (MN-DNR) ReLeaf program, administered from 1991 to

2008. This program was initiated to protect trees from oak wilt, a highly virulent disease caused by the invasive fungal pathogen Ceratocystis fagacearum (Bretz) Hunt. Grants were made available to municipal officials through this program for two specific management actions only: tree removal, and vibratory plowing, a technique to cut root grafts (Kokotovich and Zeilinger

2011). Significantly, the program did not fund fungicidal trunk injection—a treatment at that time in common use, but with little research support.

2.6 Analysis of Incentive Programs and Landowner Participation

Within forestry and extension research, the land management decisions of nonindustrial private forest (NIPF) landowners have been extensively studied, particularly in relation to incentive programs. Analyses typically focus on either assessing the biophysical outcomes of incentive payments or investigating landowner participation. Econometric methodologies are

applied to either objective, while studies of landowner participation also include those employing social psychological methodologies.

Assessment of biophysical outcomes of forestry incentive programs among NIPF landowners is typically conducted through econometric analysis of measurable market goods.

Researchers may employ a cross-sectional comparison among landowner groups (e.g., Hardie and Parks 1991) or time-series analysis of one location (e.g., Kline and Butler 2002) to examine the effect of incentives on known quantities, such as acres of forestland planted or board-feet of timber harvested. Analyses control for regionally- or temporally-varying effects such as land value, planting costs, or interest rates to determine whether financial incentives have achieved program objectives (Sun 2007). For instance, in a cross-sectional comparison of 13 southern states, Li and Zhang (2007) found a significant, positive association between NIPF acres planted and all four federal incentive programs evaluated. Biophysical outcomes of incentives for forest health treatments are more difficult to measure than those targeted at increased harvesting or planting, because of the confounding effects of weather and fluctuations in populations of pests or their natural enemies. However, Asaro, et al. (2017) noted that the recent decrease in southern pine beetle infestations correlates with federal incentives for pre-commercial thinning through the Southern Pine Beetle Prevention Program, indicating possible causal effect.

While biophysical outcomes may be measured and modeled for programs already in effect, researchers also investigate the likelihood of landowner participation in new programs, or in existing programs under different terms. Thresholds for participation are identified through contingent valuation, an economic measurement technique for non-market goods (Carson et al.

2001). Of interest are the measured values willingness-to-pay (WTP) and more commonly willingness-to-accept (WTA)—economic thresholds assigned by landowners to proposed

management actions. For instance, Kline, et al. (2000) found that payment thresholds necessary for landowners to forego timber harvest varied strongly by landownership objective. Lal, et al.

(2016) investigated per-acre payment necessary for landowners to convert non-forested land into woody biomass production, and similarly found WTA depended on current land use and acreage.

Studying Virginia’s Pine Bark Beetle Prevention Program (PBPP), a program somewhat analogous to the EABTP, Watson, et al. (2013) determined that increasing payments beyond

50% would result in only minimal participation gains.

While the goal of contingent valuation research among landowners is to identify optimal incentive payment amounts, choice modeling attempts to predict a specific landowner decision from a variety of social and geographical variables. Commonly conducted using the random utility maximization model, this technique is a special case of multinomial logistic regression in which a discrete choice is modeled as a function of several random and orthogonal parameters

(Train 1998). For instance, Joshi and Arano (2009) found the likelihood of landowners undertaking silvicultural management to be associated with educational attainment, income, forest acreage, and whether the property had been inherited. Similarly, participation in the PBPP was associated with the presence of pine on the property, education level, and the reason for landownership (Watson et al. 2013). In their study of converting non-forested land to biomass production, Lal, et al. (2016) found likelihood of planting pine to be associated with younger age, inherited land, longer ownership tenure, larger family size, and the presence of a written forest management plan.

Although random utility maximization analyses employ logit or probit models of mutually exclusive outcomes, landowner economic decisions are also modeled as a continuous functions using ordinary least squares regression. For example, in a large-scale analysis, Brooks

(1985) modeled acres of forestland planted across southern states as a function of cost-share payments, rotation length, and planting cost. In a similar study, Zhang (1996) modeled silvicultural investment of British Columbia forest landowners as a function of land tenure rights, soil quality, and biogeoclimatic zone.

Finally, social psychological theory provides an alternative perspective for the study of landowner decision-making in relation to incentive programs. Under this theoretical perspective, specific behavioral intentions are modeled as functions of measured attitudes, beliefs, and perceived norms, rather than externally-observable characteristics such as age, income, or gender. One common framework, The Theory of Reasoned Action (TRA), describes a pathway of behavioral intention that is first influenced by attitudes and subjective norms toward the behavior. Attitudes and norms are in turn informed by beliefs about behavioral outcomes and beliefs about others’ normative views (Ajzen and Fishbein 1980). Further, while external characteristics such as age or income may in fact predict or even influence behaviors, they hold no necessary relationship to the behavior (Ajzen and Fishbein 1980), and are likely to be unreliable predictors in different social settings.

Using the framework of the TRA, Sorice and Conner (2010) investigated landowner intention to enroll in a habitat conservation program, finding that attitudes and subjective norms towards participation to be the only significant predictors in a substantially-fitting model.

Similarly, in a study of participation in the NRCS’s Wetlands Reserve Program, Luzar and

Diagne (1999) found that pro-environmental attitudes predicted intention to participate although perceived peer norms did not have a significant effect. Research attention to attitudes and beliefs does not preclude consideration of external variables, however. In a study of forest landowners considering enrollment in a habitat conservation program, Farmer, et al. (2017) found that

significant predictors included both pro-environmental and family-related attitudes, along with the external characteristic of property size.

2.7 Summary of Literature

North American forests have experienced widespread economic and ecological damage from a series of invasive pests and pathogens; this trend is likely to continue. Urban forests receive a disproportionate share of this damage because of their role as ‘beachheads’ for biological invasion (Paap et al. 2017), damage that is compounded when pests find susceptible monocultures of urban trees (Alvey 2006). For these reasons, urban forest professionals and property owners have the potential to greatly reduce the severity of local and regional infestations through coordinated detection and containment of newly-arriving forest pests.

The spread of the EAB across the eastern U.S. has created more economic damage than any other invasive forest pest, and the long-term survival of eastern ash species is in question.

State forestry and natural resource agencies have responded by designing cost-share programs for municipalities that incentivize insecticidal protection of ash trees. Virginia’s EABTP also extends assistance to private property owners.

Incentive programs such as the EABTP have been developed for rural landowners by conservation and natural resource agencies since the early 1970s. Landowner participation in these programs has been intensively studied over this period. To further the conservation objectives of a program, researchers use economic and behavioral analyses to identify factors associated with the likelihood of landowner participation. There is little documentation of incentive programs designed to promote conservation objectives in urban settings; for this reason, there is a knowledge gap regarding how such a program might function. It is important to note that the actions of an urban homeowner regarding landscape trees on her property are not so

obviously an economic decision as commercial tree planting or harvest. However, since landscape trees are both a source of benefits and costs to property owners, decisions regarding their management are influenced by economic factors (Soto et al. 2018), among others.

Virginia’s EABTP provides one opportunity to study the functioning of conservation incentives among urban property owners. Because of the continuous threat posed by invasive pests to North American forests, the relevance of this research extends beyond EAB mitigation.

The EABTP provides a model for engaging private property owners in mitigating threats to forest health—threats which are posed both by newly-arriving invasive pests and other established, potentially destructive pests such as the Asian longhorned beetle.

CHAPTER 3 – SITE CHARACTERISTICS OF URBAN EABTP PARTICIPANT PROPERTIES

3.1 Introduction

In January 2018, Virginia Department of Forestry (VDOF) began offering cost-share assistance for the insecticidal protection of ash trees through its’ Emerald Ash Borer Treatment

Program (EABTP). In the program’s initial year, individual homeowners in urban areas made up

26% (28) of all participants (107). This chapter focuses on the biophysical characteristics of urban properties which received funding through the 2018 EABTP.

While the EABTP was not directed specifically towards urban homeowners, their inclusion as eligible applicants created an important opportunity for study. Since urban forests are central to the early containment of invasive forest pests (Paap et al. 2017), and most urban forests are privately-owned (Moll 1993), participation of urban homeowners in an urban forest health initiative is of research interest. To inform an assessment of the EABTP and potential implications for similar initiatives, I collected remote and on-site data for urban EABTP participant properties.

There were three specific topics addressed through this research: (1) characterization of the physical characteristics and locations of properties, (2) significance of ash species to the landscapes of properties, and (3) potential associations between the physical characteristics and the tree species composition of properties. For the first of these topics I aimed to describe urban properties and neighborhoods that benefited from EABTP funding. With the second topic, my goal was to describe the prevalence of ash trees on urban participant properties, to offer some information regarding its motivational effect. As a component of this analysis, the prevalence of ash on participant properties was also compared to that of surrounding municipalities, for added context. Lastly, by examining associations between physicial and biotic characteristics of urban

properties, I aimed to test the predictive power of site characteristics in identifying properties at high risk in a pest infestation.

3.2 Methods

3.2.1 Definition of target population

VDOF approved a total of 121 applications for the 2018 EABTP. Program policies prioritized funding of applications from the small number of Virginia counties with no detected

EAB presence in 2018 (VDOF 2018d). Ultimately, all but two approved applications were received from counties with confirmed EAB infestations (see Figure 3). Fourteen applications were eventually cancelled either by the homeowner for reasons of cost or timing of treatment, or by VDOF for reasons of poor tree condition or other disqualifying factors (Bean 2019). The remaining 107 applicants who ultimately received cost-share payments for approved ash treatment through the 2018 EABTP will be referred to here as program participants. Of these 107 program participants, 17 (16%) were public or private organizations, including municipalities, historical sites, homeowners’ associations, educational institutions, and businesses. The remaining 90 participants (84%) were single-family homeowners; 28 of these residential properties were in urban areas. I selected these properties as the target population for site inventory and analysis, referred to hereafter as urban participant properties (UPP).

3.2.2 Selection of Sites

Meetings between researchers and VDOF before the launch of EABTP in January 2018 established a data-sharing framework for this research. Since application forms were public records, most application data were shared with researchers. Application data included site

locations, number and size of ash trees, and quoted costs of treatment. However, per agreement between VDOF and researchers, contacting applicants by phone or email was limited only to those applicants who had checked an opt-in box on the program application. A 2018 EABTP application form is included in Appendix D. Addresses contained on application forms were compiled and geocoded in ArcGIS Pro 2.2 (ArcGIS Pro 2018) (hereafter ‘ArcGIS’), then joined with statewide parcel data available through Virginia Geographic Information Network (VGIN

2018a), to create a geographic database of properties of all participant homeowners.

From this database, urban properties were identified by overlaying participant locations on U.S. Census Urban Areas (U.S. Census Bureau 2018) in ArcGIS. The term Census Urban

Area (CUA) refers collectively to city-sized urbanized areas and town-sized urban clusters, both which are defined by population density and land use patterns (U.S. Census Bureau 2012).

Virginia contains 75 CUAs, irregular polygons which combine and extend beyond single- municipality jurisdictions (see Figure 3).

Of the 28 EABTP homeowner participants located in urban areas, 23 had previously opted-in to follow-up contact on the program application. Requests for permission to conduct on- site tree inventories were emailed to these homeowners in October 2018, and from this group, 17 agreed to participate. In late November and early December of 2018, tree inventory work was carried out over four days in locations between Roanoke and Alexandria. In total, 16 properties were inventoried, yielding a 57% sample of all 28 UPP.

3.2.3 Data Collection: Participant Properties

Two types of data were combined for analysis of site characteristics across UPP: compiled geographic data from public records, and on-site tree inventory data. First, five

geographic variables with relevance to forest composition were compiled from publicly- accessible sources. Parcel size was selected for its potential association with site species diversity, and household income. Parcel tree cover was selected to account for potential effects of extremely high or low tree cover on species composition. This proportional value was derived from Virginia’s Statewide Landcover Dataset, a supervised classification with 1m resolution

(VGIN 2018b). This composite dataset makes use of a number of state and federal datasets ranging published between 2005 and 2014. Historical ash abundance was included as an approximation of regional climatic or physiographic suitability for ash species. This was derived from county-level FIA data for the period 1998-2002 (USFS 2018a), collected prior to the ash borer’s arrival in Virginia. Time since EAB detection at the county level served as a proxy for

EAB population density, and was compiled from VDOF data (Chamberlin 2018b). Time since residence construction was of interest because of potential association between site species composition and distinct eras of landscape planting preferences. These data were drawn from individual searches in tax parcel databases for each respective UPP address.

To add a much greater level of detail, on-site tree inventory data were collected for 16

UPP in November and December of 2018. At each location, a complete inventory was conducted of all trees with a diameter at breast height (DBH) of 4.0 inches (10 cm) or greater. Printouts of parcel polygons overlaid on aerial imagery of each property served to indicate property boundaries in cases when property owners were not present. For each tree, species and DBH were recorded, along with five other dimensional measurements, two locational measurements, a condition rating, and a planting site descriptor. All data were initially recorded on paper before being copied to a spreadsheet. A data collection form is included in Appendix E.

Since measurements were conducted during the dormant season, tree condition rating was not based on relative canopy thinning, as is common in studies of ash tree survival (see Bick et al. 2018). Instead, I rated tree condition (0-100%) based on the relative absence of dead limbs, the relative absence of wounds or defects, and the relative presence of live buds on branch tips.

Condition ratings were collected for all tree species, as a component of a calculated tree appraisal value. Ratings were not intended as a measure of response to emamectin benzoate injection for ash trees, because treatment efficacy is not visibly evident until the growing season subsequent to treatment, at the earliest (Smitley et al. 2008).

Inventoried parcels ranged in size from 0.14 ac to 1.67 ac (0.055 ha to 0.676 ha), and in tree cover from 11.7 to 91.2%. For parcels containing unmaintained wooded areas, a single subplot of the parcel was sampled rather than the whole property. Subplots were selected to include the residence and encompass regularly maintained areas. Subplots were defined using landscape or building elements visible from aerial imagery as vertices, to aid later calculation of area in ArcGIS. This strategy was employed on 5 of 16 properties. Similarly, on 2 properties, the area inventoried was selected to intentionally exceed parcel boundaries to reflect de facto boundaries of the property as it was maintained by the homeowners.

3.2.4 Data Processing: i-Tree Eco

Before conducting further analysis, i-Tree Eco (hereafter ‘Eco’) (USFS 2018b) was used to calculate additional UPP variables from inventory data. Eco is an urban forest benefits modeling program which estimates tree benefits using allometric relationships, climate data, and measured tree dimensions. In addition to tree benefits, output includes Leaf area and Structural value. Leaf area is an allometric calculation of the total surface area of a tree’s canopy,

describing the three-dimensional effects of a tree’s canopy on a residential site. Structural value

(SV) is an estimated replacement cost, as calculated by the Council of Tree and Landscape

Appraiser’s trunk formula method (Komen and Hodel 2015).

A third Eco-derived variable used in analysis is Importance value (IV), calculated for a single tree or a species as the sum of Relative abundance and Relative leaf area. This index describes the all-around importance of a tree species to a site, combining its relative contribution to total stem count and total leaf area. For clarity, Relative importance value (RIV) is reported here, calculated as IV/2*100, and ranging from 0-100%. A table summarizing species composition variables derived from Eco calculations is included in Appendix A.

3.2.5 Data Compilation: Urban Forest Assessments

One of the objectives of this research was to investigate whether participant properties were representative of surrounding urban forests, or whether these properties were outlying cases—in effect, dense ‘stands’ of ash trees in largely ash-free urban forests. While FIA data described above provides a measure of county-level ash suitability, traditional FIA sampling does not include urban areas (USFS 2019a). Instead, data were compiled from a 2010-2011 project where Virginia Tech faculty, staff, and students conducted Urban Forest Assessments

(UFA) for five municipalities across Virginia using city-wide, plot-based techniques (Virginia

Tech 2018). UFA data were collected during a period when EAB was recently established in northern Virginia but had yet to spread outward from that region. UFA were conducted in three municipalities which surround or neighbor most UPP sites and involved plot sampling across all land uses.

UFA datasets from the 2010-2011 project include estimates and standard errors by species for stem count and structural value. From estimated tree count and structural value,

Relative abundance (RA) and Relative structural value (RSV) of all species were calculated to facilitate comparison with site data. UPP sites within 15 miles (24 km) of the geographic center of a UFA municipality were paired with that municipality. This technique paired 13 of 16 UPP properties with 3 baseline UFA datasets, and excluded 3 UPP sites from analysis. A table of site and municipal pairings with calculated distances is included in Appendix A.

Given the small sample size (n=13) and concerns about measurement precision in UFA data, the nonparametric exact sign test was used to compare species distribution metrics between properties and their surrounding urban forests. Unlike a paired t-test or a Wilcoxon signed-rank test, the exact sign test makes no distributional assumptions, comparing only the frequency of positive or negative differences between paired samples. Using this technique, the Relative abundance and Relative structural value of ash species were compared between 13 participant properties and UFA data from surrounding municipality of each propertey. Three sites were excluded from analysis because no UFA data had been collected in their vicinity.

3.2.6 Data Analysis

As a check on the external validity of site selection, inventoried and non-inventoried UPP were compared using two-sample t-tests of five geographic variables. Following these tests, a series of exploratory and inferential analyses were conducted using inventoried UPP as the experimental unit. First, as a descriptive tool, hierarchical cluster analysis was employed to define a typology of UPP by site characteristics. Secondly, the prevalence of ash trees were compared both among participant properties, and between properties and their surrounding urban forests, using

ANOVA, post hoc contrasts, and exact sign tests. Finally, ash Relative importance value and

Relative structural value were were modeled as functions of site characteristics, using multiple linear regression.

Many analyses were conducted using relative species composition indices, calculated as a tree species’ contribution to percentage of a site total. This method has the effect of giving greater weight to individual trees on sites with fewer total trees. An alternate method of comparing measurements across sites would be to standardize tree counts or basal area by parcel area inventoried. However, relative values will be employed here, because this method gives equal weight to each site, rather than individual trees. This method of calculation aligns with an analytical focus characteristics and decision-making at the household level, regardless of property size.

Two-sample t-tests, ANOVA and post-hoc contrasts were conducted in the R base package (R Core Team 2018), using RStudio (RStudio Team 2019) (hereafter ‘R’). Cluster analysis was also conducted in R, using the package ‘cluster’. Exact sign tests and multiple linear regression were conducted in SPSS version 25.0 (IBM SPSS for Windows 2017) (hereafter

‘SPSS’).

3.3 Results

3.3.1 Data Validation

Site inventories were conducted at 16 of 28 UPP properties, yielding a 57% sample. Sites that were not inventoried were omitted because of lack of homeowner interest in research participation. As a check on the representativeness of the 16-site sample, I compared means of five geographic variables between inventoried and non-inventoried properties. While there was a

marginally significant difference in tree cover between inventoried and non-inventoried sites

(p=0.053), Welch’s two-sample t-tests indicated there were no differences in means among

Parcel size, Years since home construction, Historical ash relative abundance, and Years since

EAB detection at the α = .05 level of significance. These results, summarized in Appendix A, indicate that the 16 properties inventoried provide a representative sample of the properties of all urban participant homeowners in the 2018 EABTP.

3.3.2 Cluster Analysis of Participant Properties

Averaged characteristics of all 28 UPP describe properties that are relatively large (x =

0.587 ac or 0.237 ha), wooded (mean tree cover = 47.8%), and that were developed in the late

1950s (mean age of home = 61.6 years). County-level (FIA) relative ash abundance averaged

3.59% for UPP, more than double the statewide number, and the average local age of EAB infestation was 3.89 years. Property characteristics for all 28 UPP are summarized in Table 1.

Table 1. Characteristics of all EABTP urban participant properties (n = 28). x̅ SD Parcel size (acres) 0.587 0.106 Parcel tree cover (%) 0.475 0.053 Years since home construction 61.6 7.9 Historical ash relative abundance1 (%) 3.59 0.53 Years since EAB detection 3.9 0.7 1Summarized at the county level from pre-EAB FIA data (1998-2002)

To help characterize the types of neighborhoods which benefitted from EABTP, I developed a a typology of inventoried urban participant properties. Typologies are a common tool in landowner research, used to segment groups with minimal loss of information (Dayer et al. 2014). Using the R package ‘cluster’ I conducted a hierarchical cluster analysis using normalized values of the variables Parcel size, Parcel tree cover, and Years since home construction. This analysis met the minimal sample size threshold of 2k, where k is the number of variables (Dolnicar 2002). From the resulting dendrogram (Figure 4), I selected a 3-cluster solution as the most interpretable, yielding two clusters of five properties each, and one of six.

Mean values for property characteristics used in analysis are shown by cluster in Table 2, which includes additional site and species composition characteristics.

Cluster analysis yielded a meaningful typology, based only on the physical characteristics of parcel size, tree cover, and age of the home. Properties in Cluster 1 were largest, with a mean area of 1.00 acres (0.40 hectares), had the highest tree cover (x̅ = 80.8%) and were developed in the mid-20th century (mean age = 42.8 years). These characteristics describe large, wooded exurban properties, often built into forested areas, and are termed here Wooded exurban.

Associated species composition data (Table 2) indicated that properties in Cluster 1 had, on average, the highest total basal area (x̅ = 60.2 ft2) for all trees, but the lowest average relative basal area (x̅ = 27.0%) for ash species.

Properties in Cluster 2, were smaller (mean area = 0.38 ac or 0.15 hectares), had a relatively lower amount of tree cover (43.2%) and were developed pre-WWII (mean age of house = 100.8 years). These characteristics include landscaped properties of urban neighborhoods developed in the 1930s and before, termed here Historic urban. On these

properties, both the relative abundance (x̅ = 19.6%) and relative structural value (x̅ = 43.0%) of ash species were higher than in the other clusters (Table 2).

Finally, characteristics of properties in Cluster 3 fell mostly between those of the other two—size and tree cover was moderate (mean area = 0.47 ac or 0.19 hectares, mean tree cover =

41.5%), as was mean age of development (25.8 years). These characteristics describe suburban development of recent decades, termed here Contemporary suburban. This typology of properties, in addition to its use as a descriptive tool, contributed categorical predictors to models of site species composition. On average, these properties had the lowest total basal area for all trees (x̅ = 28.2 ft2), and the lowest relative abundance of ash species (x̅ = 13.2%) (Table 2).

Figure 4. Hierarchical cluster analysis dendrogram of Urban Participant Properties (n=16), calculated from a distance matrix of Parcel tree cover, Parcel size, and Years since home construction. The cut-off point for a three cluster solution is represented by the red dotted line.

Table 2. Property characteristics by cluster membership for 16 inventoried Urban Participant Properties. Mean values are shown. Cluster 1 2 3 Wooded Historic Contemporary exurban urban suburban Parcel size (acres) 1.00 0.38 0.47 Tree cover (%) 80.8 43.2 41.5 Years since home construction 42.8 100.8 25.8 Historical ash relative abundance (%) 2.8 3.8 2.9 (county-level FIA data) Years since local EAB detection 7.0 2.8 2.8 Number of trees on property (> 4" DBH) 33.0 15.8 21.0 Total basal area (ft2) 60.2 28.7 28.2 Ash relative abundance (%) 17.0 19.6 13.2 Ash relative basal area (%) 27.0 43.0 31.4

3.3.3 Comparison of Species Composition Among Sites

Across all 16 inventoried UPP, white ash (Fraxinus americana) topped most measures of species composition. In absolute terms, as measured by total tree count (59) or total basal area

(234 ft2 or 21.8 m2), white ash surpassed all other species by a factor of more than four.

Similarly, white ash ranked higher than all others species by relative abundance, indicating that on average, it was the most common tree on each site. An ANOVA and post-hoc Tukey’s HSD test demonstrated a statistically significant difference in relative abundance between white ash and all other species, with the exception of red maple (Table 3).

As measured by relative structural value—an appraisal value calculated from trunk diameter, condition, and species—white ash was also the highest-ranked species. In this case, there was a statistically signicant difference between white ash and all other species (Table 3), demonstrating that on average, this species provided greatest contribution to property value of all

tree species recorded. Figures 5 and 6 display the distributions of inventoried species by relative measures across sites. Notably, while white ash represented 16% of mean relative abundance, green ash (Fraxinus pennsylvanica) was present on only one site, with a mean relative abundance of 0.27%.

As with species composition, F. americana heavily contributed to calculated tree benefits across all sites. Tree benefits calculated by Eco include monetized values for total and annual carbon storage, runoff avoided, pollution removal, energy savings, and carbon emissions avoided. For F. americana, the summed monetary value of these five benefits, averaged across all sites, was nearly 6 times greater than for Liriodendron tulipifera L., the next largest contributor. Summaries of tree benefits, site characteristics, and species composition indices are included in Appendix A.

3.3.4 Comparison of Species Composition Between Sites and Surrounding Urban Forests

At a glance, species composition figures drawn from municipal UFA data appeared noticeably different from data collected at sites. Species with highest mean relative abundance for Roanoke, Charlottesville, and Falls Church included both the invasive species, Ailanthus altissima (Mill.) Swingle, and the ornamental Cornus florida L. Species rankings by structural value, on the other hand, were dominated by overstory trees including Liriodendron tulipifera and Liquidambar styraciflua L. Figures 7 and 8 show the distributions of species by relative abundance and relative structural value, averaged across all three UFA municipalities.

Results from exact sign tests indicated that both ash relative abundance and relative structural value were significantly higher on participant properties than in surrounding urban forests. In the case of relative abundance of ash, values were higher for all 13 properties relative

to municipalities, with a statistically significant median decrease of 10.4% (p <.001). In the case of relative structural value of ash species, values were higher for 12 of 13 sites, with a statistically significant median difference of 26.3% (p = 0.003) (Table 4).

Table 3. ANOVA comparison of the top ten species by relative abundance and relative structural value across 16 inventoried urban participant properties. Mean values are expressed as a percent of site totals. Relative abundance Relative structural value Species x̅ Group1 Species x̅ Group1 Fraxinus americana 16.59 a Fraxinus americana 31.38 a Acer rubrum 8.55 ab Liriodendron tulipifera 9.04 b Celtis occidentalis 4.92 b Acer rubrum 7.70 b Cornus florida 4.36 b Acer saccharinum 5.79 b Acer saccharinum 4.22 b Quercus rubra 4.44 b Morus alba 3.72 b Quercus alba 3.34 b Liriodendron tulipifera 3.68 b Quercus palustris 3.10 b Pinus strobus 3.59 b Morus alba 2.60 b Juglans nigra 3.16 b Juglans nigra 2.55 b Magnolia grandiflora 2.80 b Celtis occidentalis 2.44 b F 2.76 F 8.67 p 0.01 p <.001 Bolded values indicate significance difference in means at the α = .05 level of significance. 1Group labels display results of post-hoc Tukey's HSD tests, indicating significant differences by letter.

Table 4. Results from exact sign test of species composition between 13 Urban Participant Properties and surrounding urban forests. Properties were paired with nearby or surrounding municipalities; median differences and Z-statistics reflect the aggregated differences between the two. Participant Surrounding Median Dependent variable Z p properties urban forests difference Ash relative abundance 14.3% 3.9% 10.4% -3.33 <0.001 (median) Ash relative structural 30.8% 10.4% 26.3% -2.77 0.003 value (median) Bolded values indicate significance at the α=0.05 level.

Figure 5. Relative abundance of species across 16 EABTP Urban Figure 6. Relative structural value of species across 16 EABTP Participant Properties. Top ten species by relative abundance Urban Participant Properties. Top ten species by relative structural are shown. value are shown.

Figure 7. Relative abundance of species across 3 Virginia Figure 8. Relative structural value of species across 3 municipalities. Data were drawn from 2010-2011 Urban Forest Virginia municipalities. Data were drawn from 2010-2011 Assessments of Falls Church, Charlottesville, and Roanoke. Top Urban Forest Assessments of Falls Church, Charlottesville, ten species by relative abundance are shown, with the addition of and Roanoke. Top ten species by relative abundance are F. pennsylvanica and F. americana. shown, with the addition of F. americana.

3.3.5 Linear Models of Species Composition

Following cluster analysis of property characteristics, I modeled species composition across properties using multiple linear regression in SPSS. Of the four ash-related variables measured or derived from inventory data (Relative abundance, Relative basal area, Relative importance value, and Relative structural value), I selected the two containing the most information as dependent variables: Relative importance and Relative structural value. The index

Relative importance combines both species abundance (tree count) and leaf area in a single measure, while the index Relative structural value incorporates trunk diameter, a measure quadratically proportional to basal area, in addition to tree condition and species rating.

With the intention of building predictive models of site species composition that could be replicated without on-the-ground data collection, I selected only publicly-accessible predictors.

Those entered in the models were cluster membership, Years since EAB detection, and Historical ash abundance. Cluster membership summarized the variables of Parcel size, Parcel tree cover, and Years since home construction, and was entered in the model using dummy variables.

Cluster 2 was left out of models as a reference level—with the smallest lot size, least tree cover, and oldest age of home construction, it provided the most interpretable point of comparison.

Models of ash relative abundance and relative structural value met assumptions for independence of observations, linearity, homoscedasticity of variance, collinearity, normality, and lack of influential points. However, testing indicated one outlier and nine points with high leverage in each model. The presence of several unusual points—which were not removed from the model—necessarily reduced model fit and indicate the need for a greater sample size in future analyses.

The model of ash Relative importance value fit poorly (R2=1.97, Adj. R2 = -0.095), was not significant overall (F = 0.674, p = .623), and did not contain any individually significant predictors (Table 5). A negative value for adjusted R2, which is corrected for the number of parameters in the model, indicates that a model using the constant alone fits the data better than the full model. Modeled ash Relative structural value fit slightly better (R2 = 0.402, Adj. R2 =

0.185), although it too showed no significance overall (F=1.85, p = .190) and contained no significant predictors. The only variable of marginal significance to Relative structural value was

Years since local EAB infestation (ß= -0.663, p = .093).

Table 5. Summary of modeled ash importance and structural value across 16 urban participant properties in the 2018 EABTP. Model of ash Model of ash Relative Relative importance value structural value Independent Variables 훽 coeff. p 훽 coeff. p (Constant) -- 0.010 -- 0.001 Years since local EAB detection -0.544 0.220 -0.663 0.093 Historical ash abundance (FIA) -0.331 0.339 -0.314 0.296 Cluster 1 - Wooded exurban 0.033 0.933 -0.134 0.692 Cluster 2 - Historic urban ------Cluster 3 - Contemporary suburban -0.317 0.330 -0.461 0.114 R2 0.197 0.402 2 Adjusted R -0.095 0.185 F 0.674 0.623 1.850 0.190

3.4 Discussion

While broadly generalizable conclusions are unlikely from a small sample size, cluster analysis provided some insight into the locations where 2018 EABTP funding was utilized.

Participant properties were split in roughly equal numbers between large, exurban wooded properties, smaller, urban (and older) properties, and suburban properties with characteristics roughly in between those of the others. This distribution of participant properties is not necessarily intuitive—one might expect a greater share of smaller participant properties with a single, specimen ash tree, and correspondingly low tree cover on a smaller lot. Instead, most urban-area residents interested in EABTP appear to be located on the suburban and exurban periphery, on parcels with relatively high tree cover—some likely built into secondary-growth forests.

Tests for differences in means of species composition indices across inventoried sites demonstrated the consistent dominance of white ash on these properties. With the exception of red maple, the relative abundance and structural value of white ash were significantly greater than all other species. If the opposite had been true—and ash trees were present in roughly equal number to several other species—it might suggest that tree preservation was driven primarily by appreciation of individual trees. However, the observed results indicate an association between the relative importance of ash on a residential property and the homeowner’s interest in cost- share participation.

Exact sign tests confirmed that the relative abundance and value of ash trees were greater on participant properties sites than in the surrounding cities. These results must be qualified by noting the age and standard errors of UFA measurements (summarized in Appendix A), and the low sample size (n = 13) of properties compared. However, the age of the UFA datasets may in

fact serve to support the test results. UFA data were collected from 2010-2011, before the ash borer was established outside of northern Virginia. During this period, EAB was present in Falls

Church, but likely did not reach Roanoke until 2016 and Charlottesville in 2017. It is noteworthy that relative ash importance and value were higher in 2018 on participant properties than compared to citywide pre-EAB figures, in the case of Roanoke and Charlottesville. These results again suggest that properties where ash preservation takes place are likely to be outliers relative to neighboring properties in terms of the number and value of threatened trees.

Because of the small sample size, only tentative conclusions could be inferred from linear modeling results, even if models had fit the data closely. The one marginally significant association was between the predictor Years since EAB infestation and the dependent variable

Ash relative structural value. This negative association is intuitive—as time passes since a local initial EAB detection, the number and condition of ash trees decline, leading to a decrease in value. Interestingly, the same predictor was not associated with Relative ash abundance, which only accounts for the number ash trees, and not their condition. Since properties had, on average, experienced about four years of a local EAB infestation, it is likely that a sharp drop in relative abundance of ash trees might be apparent if this analysis were repeated one or two years later, as greater mortality of infested trees would likely be evident. A broader discussion of this topic and other study findings, implications, and directions for further study are included in Chapter 6.

CHAPTER 4 - HOMEOWNER ENGAGEMENT IN LANDSCAPE TREE PRESERVATION

4.1 Introduction

Financial incentive programs have been employed by U.S. agencies for decades to further soil conservation, wildlife, or forestry objectives among rural landowners, yet have rarely been employed in urban areas. However, among researchers and practitioners there is a growing focus on the health and environmental benefits provided by landscape trees and other urban natural resources (Kondo et al. 2018). Considering that over 80% of the U.S. population now resides in urban areas (USFS 2019b), it is likely the demand for public funding of urban conservation efforts will grow.

While the relatively small parcels of suburban and urban development could be viewed as a hindrance to the widespread impact of an incentive program, there may also be advantages unique to funding conservation work in cities. Compared with agricultural or forested land, the number of property owners with a personal stake in local conservation actions is many times greater in urban areas, when calculated by tree or by acre. Because of cities’ density, incentive programs aimed at enlisting urban homeowners in forest health initiatives have the potential to encourage widespread engagement in pest detection and strongly augment public investment in pest management. While rare, successful eradication campaigns such as those targeting Asian longhorned beetle in Chicago, IL, Worcerster, MA, and Jersey City (Haack et al. 2010) have been accomplished through rapid, costly, and coordinated actions in urban forests.

Virginia Department of Forestry’s (VDOF) Emerald Ash Borer Treatment Program

(EABTP) is one of few incentive programs to include urban property owners among eligible participants. Consequently, this program provided an important opportunity for study. In its

initial year, this program provided 50% cost-share payment for treatment of ash trees to 90 private property owners, of which 28 were in urban areas. Regardless of location, EABTP funding was used for insecticidal treatment of individual landscape trees, rather than silvicultural treatment of a forest stand as might be conducted in a traditional cost-share program. In this regard, the experience of rural participants was similar to that of urban participants.

To examine the factors influencing a homeowner’s participation in a forest health cost- share program, surveys were distributed to homeowners drawn from two distinct target populations. The first consisted of all homeowner participants in the 2018 EABTP (hereafter,

‘program participants’). The second was defined loosely as shade-tree-owning Virginia homeowners in three urban study areas (hereafter, ‘general households’). The survey instruments designed for each group shared most survey items, which allowed for comparisons between program participants and general households, who could be viewed as potential participants.

Stated most broadly, the objective of this research was to help inform future urban forest health initiatives. More specifically, this research addressed questions regarding homeowners’ willingness to pay for the preservation of threatened trees, homeowners’ interest in cost-share funding, and the difference in attitudes between program participants and a sample of urban households. By examining perceptions among both groups, this research studied the feasibility of an EABTP-like program among a broad group of urban homeowners, in addition to studying the first-year implementation of the EABTP itself.

The theoretical framework adopted here combined elements of the Theory of Reasoned

Action (TRA)—focused on the predictive power of specific attitudes and underlying beliefs

(Ajzen and Fishbein 1980)—and economic choice modeling, which employs externally- measurable variables as model parameters. This strategy reflected the research objective of

informing urban forest health initiatives, by identifying information with greatest utility to a hypothetical program manager—whether drawn from public databases or opinion surveys.

4.2 Methods

4.2.1 Construction of Survey Sampling Frames

For all homeowner survey research, the individual household was considered to be the survey element. Survey items did request ownership status, age, and gender, but did instruct recipients regarding which member of the household should complete the survey. All survey recruitment materials and instruments were submitted first to the Virginia Tech IRB and subsequently approved by Western IRB. An IRB Exemption letter is included in Appendix H.

Program Participants

Homeowners accounted for 90 of 107 total participants in the 2018 EABTP. Of these, 28 were in urban locations, as defined by Census Urban Areas (CUAs). All 90 program participants, including those in rural and urban locations, were considered to be the target population. While the use of unique survey links allowed for disaggregation of response data by location, analysis was conducted with aggregated urban and rural data. This decision reflected an assumption, supported by later analysis, that homeowner decision-making regarding preservation of individual landscape trees would not greatly differ between rural and urban locations. Survey requests were sent only to 77 of 90 program participants—those who had previously opted-in to follow-up contact when applying to the EABTP. For the purposes of calculating response rate and examining potential nonresponse bias, however, the entire population of 90 homeowner particpants was considered to have been part of a complete sample.

General households

Construction of a sampling frame of Virginia urban homeowners with no known connection to the EABTP was accomplished using publicly-available geographic and tax assessment data. Primary household selection criteria included (1) urban location, (2) owner occupancy, and (3) likely shade tree ownership. Taken together, criteria were intended to select households with a personal stake in urban tree preservation—those whose attitudes and actions were directly relevant to urban forest health. The three primary criteria were in turn defined by a series of filtering steps conducted in ArcGIS Pro. Lists of data sources used in each step of this process are summarized in Appendix B .

(1) Urban location

Virginia counties and independent cities were initially stratified by time elapsed since EAB

infestation. Strata definitions were borrowed from VDOF criteria establishing EABTP

funding priority (VDOF 2018b): counties with no detected EAB presence (undetected),

counties where EAB was detected in 2015 or later (recent), and counties where EAB was

detected before 2015 (established). From within each EAB stratum, the boundaries of the

most populous Census Urban Area (CUA) were selected to define study areas, then further

narrowed to selected contiguous city and county jurisdictions. Limiting household sampling

to a total of eight jurisdictions increased efficiency of data compilation and bulk mailing, as

compared to sampling throughout an entire CUA. The resulting study areas, referred to here

as Elizabeth River, Roanoke Valley, and Northern Virginia, correspond with the EAB strata

of undetected, recent, and established, respectively. Figure 9 displays Virginia counties by

EAB strata, CUAs, and study areas.

(2) Owner occupancy

Parcels within each study area were filtered to identify households with sole, direct

responsibility for property maintenance. For this reason, selection criteria included only

owner-occupied, single-family residences, and excluded rented homes, condominiums, or

other multi-unit parcels.

(3) Shade tree ownership

Selected owner-occupied parcels with single family residences were further filtered to

include only those with a high probability of shade tree ownership, using criteria for

minimum area (0.2 ac) and minimum tree cover (25%). Together, these ensured an absolute

minimum of tree cover and a minimum relative to parcel size. This step was intended to

improve the probability of selecting homeowners with direct ownership of multiple

landscape trees and accustomed to the responsibilities of landscape tree maintenance.

From parcel lists aggregated by study area, 500 records from each were selected at random to form a mailing list. Address validation removed 12 records, resulting in a final mailing list of 1,488 households. Table 6 outlines sampling frames and mailing list totals for each study area. Since the number of households in each sampling frame were not equal, the probability of selection was higher for households in study areas with fewer members. Mailing lists were not adjusted for proportionalilty to population for two reasons: (1) study area boundaries themselves did not reflect regional population distribution, and (2) roughly equal numbers of of survey responses were desired from each study area, to facilitate comparisons.

Figure 9. Map of EAB infestation strata, Census Urban Areas, and study areas for the general household survey. Infestation strata were defined by the year of initial EAB detection in each Virginia county or independent city.

Table 6. Total population, sampling frame size, and number of survey recipients, by study area for general household survey. Number of households Survey Sampling Total meeting recipients Study area Jurisdictions frame by population1 criteria by study study area (Sampling area frame)

Northern Arlington County 226,092 7,775 Virginia Alexandria City 151,473 2,045 10,964 495 Falls Church City 23,620 1,144 Roanoke Valley Roanoke City 99,329 5,651 Roanoke County 93,655 6,087 12,806 497 Salem City 25,290 1,068 Elizabeth River Chesapeake City 233,194 7,039 10,676 496 Portsmouth City 96,071 3,637 Total 948,724 34,446 34,446 1,488 1 U.S. Census Bureau (2017a)

4.2.2 Construction and Pre-testing of Survey Instruments

While survey instruments for participants and general households were distributed by different methods, both were constructed with Qualtrics (Qualtrics 2018), employed the same structure and shared most survey items. The most important difference between the two survey instruments was that participants homeowners were asked about their experience with the

EABTP, while general households were asked about interest in a hypothetical cost-share program for landscape tree preservation. With the exception of exploratory factor analyses, data from the two surveys were not aggregated for analysis.

Program participants

The survey instrument designed for participants consisted of 28 questions, with an estimated response time of 12 minutes. Survey items consisted primarily of multiple-choice questions, including one multi-item index, and a single ranked-choice response item. A summary of survey items and measurement scales is given in Appendix B, while the full text of this survey instrument is available in Appendix G.

General households

The survey instrument designed for general households consisted of 26 questions, with an estimated response time of 15 minutes. Wording of survey items was similar to those used for the participant survey but was modified to reflect two assumptions: first, that recipients would likely not be aware of the EABTP, and secondly, that most were unlikely to have ash trees on their properties. This second assumption was based on Virginia Urban Forest Assessment data, discussed in Chapter 3, indicating pre-EAB relative abundance of ash species did not exceed 4%

in Virginia municipalities. For this reason, the wording of survey items discussed “landscape tree preservation” and “threatened trees” rather than ash preservation or EAB specifically. The full text of this survey instrument is available in Appendix G.

Pre-test of Survey Instruments

To assess clarity and measurement accuracy of survey questions, a combined pre-test for homeowner survey instruments was conducted in August 2018, using the wording of the participant version. This pre-test survey instrument included all proposed survey items and one text-entry response item, asking for respondents’ comments regarding length, wording and logical flow. Pre-test survey requests were emailed to Virginia Cooperative Extension (VCE) agents working in urban areas (n = 50). This group was selected as a pre-test sample under the assumptions that many would be familiar with invasive species management, and likely would own homes within the urban areas they served. Seventeen recorded responses assisted in the improvement of several ambiguously-worded questions, and in the re-ordering of a measurement scale which initially produced inaccurate results.

4.2.3 Data Collection

Program participants A web survey was selected as an efficient and appropriate mode to reach participants.

Because current email addresses were uniformly available for all members of the sampling frame, the potential for technological response bias was minimized (Vaske 2008). Survey requests were emailed to participants, while the web survey itself was hosted by Qualtrics. In

September 2018, ‘opt-in’ participants (n=77) were contacted by email with an introductory survey request. Building on elements of Dillman’s Tailored Design Method (Dillman et al.

2009), up to two reminder emails were sent with varied content to nonrespondents, separated by periods of ten days. Survey requests contained unique survey URLs, which allowed tracking of recipients’ response status, and later joining of response data with geographic variables, such as tree cover. Recruitment materials and survey instrument for the participant survey are available in Appendix F. Response data were recorded and stored by Qualtrics during the period the survey was available online. The participant survey was closed at the end of December 2018, about three months after homeowners were initially contacted.

General households

A mixed-mode web+mail survey was selected to reach general households in the study areas of Northern Virginia, Roanoke Valley, and Elizabeth River. Mixed-mode surveys have the potential to compensate for weaknesses of web- or mail-only designs by offering recipients the choice of their preferred mode of response (Vaske 2008). Other advantages of a web+mail survey design over a mail-only design include lower costs, reduced response times, and potentially improved response rates (Dillman et al. 2009). While some researchers studying web+mail surveys have noted reduced response rates compared to mail-only surveys (Medway and Fulton 2012), others have documented an upward trend in the proportion of web respondents over time (Lesser and Newton 2016). Studies employing a web+mail survey design have also reported that younger age groups show preference towards an online response option (Sexton,

Miller, and Dietsch 2011, de Bernardo and Curtis 2013). These findings suggested that a web+mail design would be an appropriate tool for reaching urban areas, where the adult age distribution is younger than that of rural residents (U.S. Census Bureau 2016). This assumption was later examined in an analysis of response mode.

Web and mail versions of the survey instrument were created in Qualtrics and in booklet form. To allow tracking of individual responses, identification codes were assigned to each of

1488 households. This unique code was printed on each recruitment letter and survey booklet; those responding to the web version of the survey were asked to enter the code at the beginning of the questionnaire. Except for this initial item which did not appear in the survey booklet, there were no differences in the wording or order of questions between web and mail survey versions.

In September 2018, general households were contacted by mail with an introductory letter informing them about this research project and requesting their participation. The introductory letter outlined two options for those interested in completing the survey: waiting to receive a survey booklet through the mail or completing the same survey online. The letter also supplied instructions and a URL for the web option. Two weeks following the mailing of the introductory letter, survey packets were mailed to all homeowners (n=1,488), containing a cover letter, a survey booklet, and pre-paid business reply envelope. Like the introductory letter, the enclosed cover letter provided a URL and instructions for recipients who preferred to complete the survey online. After compiling early web and mail responses, a third and final letter was mailed to nonrespondents (n=1,136) two weeks after mailing the survey packets. This letter requested participation, provided instructions for online survey completion, and offered to send a replacement survey packet upon request. The survey instrument and recruitment materials for the general household survey are available in Appendices F and G.

Web survey response data were recorded and stored by Qualtrics during the period the survey was available online. At the end of December 2018, the general household web survey was closed—about three months after homeowners were initially contacted. Survey booklet

responses received by mail were manually entered in Qualtrics with postmarked dates individually recorded.

4.2.4 Data Analysis

Recoding of Binned Data

Before other analyses were conduted, binned survey and Census data which represented underlying continuous data were recoded as continuous variable, using the R package

‘binequality.’ Bin midpoints were calculated as the mean value of upper and lower bin bounds, and then substituted as the recorded value. For instance, a survey response of $50,001 to $75,000 for household income was interpreted as the mean value of $62,500.50. For uppermost bins with no upper bound, a pseudo-midpoint was estimated, using a Pareto distribution fitted to data from the two highest bins (von Hippel et al. 2017). For measures of household income, this function was fitted with an alpha value of 1.11, approximating a traditional Pareto wealth distribution. For all other binned variables with undefined upper bounds, this function was fitted with an alpha value of 111, to produce much more conservative estimates. This technique improved interpretability of results, and potentially helped re-create a more accurate distribution of data.

Aside from household income, this technique was used calculate midpoints and pseudo- midpoints for variables including Age, Age of home, Number of years living at property, Number of trees on property, and Years of education.

Tests for Nonresponse Bias

To determine whether survey response data provided a basis for generalizable conclusions, tests of nonresponse bias were conducted by comparing means of external variables between respondent and nonrespondent groups. External variables included two parcel-level

property characteristics and four demographic characteristics summarized at the level of Census

Block Groups. For parcel-level comparisons, Parcel size and Parcel tree cover were calculated in ArcGIS from the Virginia Parcels geodatabase (VGIN 2018a) and the Virginia Statewide

Land Cover Database (VGIN 2018b).

For Block Group-level comparisons, Census-derived estimates served as geographic proxies for households contained within them. Testing of proxy variables, including aggregate

Census data, has increasingly been used by researchers to assess and correct for survey nonresponse bias (Biemer and Peytchev 2013), and is an efficient option when follow-up contact with nonrespondents is not feasible (Hansen et al. 2007).

Using the U.S. Census Bureau’s American Factfinder (U.S. Census Bureau 2017a), I accessed 2017 American Community Survey 5-year estimates of demographic variables for

Virginia Block Groups of all respondents and nonrespondents to either survey. Census Block

Groups are the smallest U.S. Census geographic divisions, drawn to include 600 to 3,000 people

(U.S. Census Bureau 2010). The tables I selected for analysis were Median Age by Sex

(B01002), Median Household Income in the Past 12 Months (B19013), Educational Attainment for the Population 25 Years and Older (B15003), and Race (B02001). Households in question were joined to Block Group estimates, creating a many-to-one relationship where several households might share the same Block Group estimate. Frequency data from the Census table

Race were used to calculate a proportional variable Minority Proportion, defined as the proportion of all residents who did not identify as ‘White alone.’

Continuous proxy variables Parcel size, Mean age, Median household income, and Mean years of education were compared using Welch’s t-test, while proportional variables Parcel tree cover and Minority proportion were compared using Pearson’s χ2. Because repeated testing of

the same dependent variable (respondent group) raised the probability of Type I error, a

Bonferroni correction was applied to test results, lowering the significance level to α = 0.008.

Factor Analysis of Attitudes Towards Urban Trees

To reduce the number of predictor variables and examine potential latent structures

(Yong and Pearce 2013), an Exploratory Factor Analysis (EFA) was conducted of five attitudinal survey items measure on identical 5-point Likert scales. In these survey items, respondents were asked to rate the strength of their agreement with five statements regarding the importance of urban landscape trees. For this analysis only, responses were aggregated between participants and general households, to facilitate cross-sectional comparison. Analysis was conducted using the R packages ‘pysch’ and ‘GPArotation,’ with varimax rotation selected. Two factors were retained, as indicated by the inflection point of eigenvalues in a scree plot (Costello and Osborne

2005).

Cluster Analysis of Ranked Motivations for Tree Preservation Cluster analysis, like factor analysis, is typically employed as a dimension-reducing technique. In this case, cluster analysis was used to accomplish the opposite: from a single item on each survey, a categorical typology of homeowners was developed. The survey item in questions asked respondents to rank six motivations for ash tree (or shade tree) preservation in order of personal importance. Response data collected from this item consisted of an ordered vector of numbers 1-6, representing six ranked motivations, including a write-in option. This form of response data is potentially information-rich—examples of applications for clustered ranking data include political affiliation patterns in voting records (Gormley and Murphy 2008), or market segmentation by brand preference (Müllensiefen et al. 2018).

Ranking of motivation for preserving trees intended to measure the degree to which intrinsic or extrinsic motivation informed respondents’ attitudes. Researchers have theorized that intrinsic motivations—reasons for action based on personal drive or interest—are potentially more stable than extrinsic motivations, such as financial gain (Ryan and Deci 2000). Of the five options given, intrinsic motivations were represented by the provision of wildlife habitat and prevention of species endangerment, while extrinsic motivations were represented by the importance of trees as attractive landscape elements or their contribution to property value. The fifth motivation listed—trees’ provision of shade—did not clearly fit in either category.

Ranking data were analyzed using the R package ‘Rankcluster.’ The sorting algorithm employed measures distance from each ordered response to a given number of modal ranking patterns, forming clusters based on relative similarity (Jacques, Grimonprez, and Biernacki

2016). The optimal number of clusters was determined from an inflection point of BIC plotted against the number of clusters, similar to the technique used in factor analysis. A 2-cluster solution was selected for participants, and 5-cluster solution for general households.

Linear Models of Behavioral Intention

Homeowner engagement in urban forest health was examined through linear models of two dependent, behavioral variables. These were respondents’ level of Tree preservation intention—meaning their willingness to pay for ongoing tree treatment—and their level of interest in Cost-share participation. Survey item wording varied between the two survey instruments: program participants were asked specifically about ash tree preservation and the

EABTP; general households were asked about the more generic topics of landscape tree preservation, and participation in a hypothetical program for assistance with tree treatment

expense. A summary of all survey questions is included in Appendix B, while the full survey instrument is included in Appendix G.

Models were constructed of predictor variables drawn from external geographic data and survey response data, including both factor indices and cluster membership, which were derived from survey response data. The full lists of predictor variables—which totaled 17 for participants and 25 for general households—were categorized into one of three predictor classes, each consisting of 4 to 8 individual predictors: Property characteristics, Personal characteristics, and

Attitudes. In addition to aiding interpretability of results, the three predictor classes as ordered here represent an increasing difficulty of access, from the point of view of a public official. A full list of all predictor variables, organized by class, is included in Appendix B.

Model selection was accomplished by ranking of candidate models in order of decreasing

Akaike’s weight (wi), a proportional measure of relative likelihood derived from AIC. This method of model selection relies on an information-theoretic framework in which multiple competing models are evaluated in terms of information loss, as measured by AIC (Symonds and Moussalli 2011). Models within 2 AICc (a version of AIC corrected for small samples) of top-ranking models were examined as competitors to top-ranked models if they contributed no more than one additional parameter, as recommended by Arnold (2010).

Candidate models were constructed from the predictor classes of Property

Characteristics, Personal Characteristics, and Attitudes. For both dependent variables in question—Tree preservation intention and Cost-share participation—a total of seven candidate models were fitted, constructed from all combinations of individual, paired, and grouped predictor classes. This process was conducted for both participant and general household survey

response data, resulting in a total of four top-ranked models. All modeling was carried out using hierarchical multiple linear regression in SPSS.

4.3 Results

4.3.1 Summary of Response Data

Program participants Over a period of about three months, 53 responses were recorded, representing a 69% of response rate out of 77 participants contacted. With the inclusion of 13 additional ‘opt-out’

EABTP homeowner participants in the sampling frame, this is equivalent to a 58% response rate overall. Calculated from the complete sampling frame, the margin of error for a 95% confidence level was 8.63%. This value represents the sampling error associated with a hypothetical 50/50 split response, calculated from sample size and final sample (number of responses) (Vaske

2008).

Among participant respondents whose locations were known, there were twice as many from rural areas (n =34) as from urban areas (n =17). On average, respondent properties were large (x̅ = 30.62 ac or 12.39 ha) and had high tree cover (x̅ = 58.6%). Length of time since local

EAB detection varied between respondents from 0 (undetected) to 10 years (established), with a mean value of 2.4 years.

On average, respondents were nearing retirement (mean age = 59), had completed a graduate degree (x̅ = 19.0 years of education), and were predominantly male (60.3%). All respondents who disclosed racial or ethnic identity (n = 50) identified as White. Average annual household income was $232,000, although the variability associated with this estimate was high

(SD = $160,000). Respondents reported living on their properties for an average of 11.3 years, and a majority (56.6%) reported previously paying for insecticidal treatment of landscape trees.

Between maintained and unmaintained areas, respondents reported having over 20 trees on their property (x̅ = 21.4) and spending close to $1000 annually on tree maintenance (x̅ = $987).

A plurality reported first hearing about the EABTP from an arborist or landscape professional (n=13), followed by VDOF or VCE employees (n=12), then coworkers or friends

(n=12). Asked about the importance of preserving ash trees in Virginia, participants’ mean response was 4.4 on a scale from 1 (Not at all important) to 5 (Extremely important). When asked about the likelihood of regularly paying for ash tree treatment going forward, respondents’ mean score was 4.6 on the same scale. Finally, when asked about interest in re-applying for

EABTP funding, the mean score was also 4.6 on a scale from 1 (Not at all interested) to 5

(Extremely interested). Table 7 summarizes these characteristics alongside those of general households. Additional attitudinal response data are discussed below with exploratory and cluster analyses.

General households

More responses were received by mail (n=206) than online (n=142). After removing 15 blank or incomplete responses, a total of 333 usable responses remained. Two responses were excluded from analysis because respondents indicated they were renting their home, and thus did not meet sampling frame criteria. An additional 10 responses representing ash tree-owning households were excluded from primary analysis, because these respondents were asked alternate versions of many survey items. After removing unusable and incompatible responses, a total of 321 responses remained, for a final response rate of 21.6%. While this response rate was considerably lower than among participants, the calculated margin of error (4.84%) was smaller than that of the participant survey due to the larger sample.

In contrast to the predominantly rural participants, all respondents to the general household survey lived in urban areas. Respondents’ average parcel size was 0.442 acres (0.18 ha), a fraction of the amount for participants, but still relatively large for urban areas. Average tree cover across all respondent parcels was 45.0%—well above the 25% threshold used in sampling frame construction. Of the three study areas targeted, 125 (38.9%) responses were received from Northern Virginia, 116 (36.1%) from Roanoke Valley, and 80 (24.9%) from

Elizabeth River. As of 2018, EAB infestations had been present in these these study areas for 10,

2, and 0 years of EAB respectively. Across all responses, the average length of time since EAB detection was 4.7 years, about double that of participants.

Demographic characteristics of general households were similar to those of participants.

On average, respondents were also approaching retirement age (x̅ = 58.7 years) and had completed a four-year degree (x̅ = 17.5 years of education). A majority of respondents were male

(53.6%), although the overall ratio of males to females was more balanced than among participants. Of respondents who disclosed racial or ethnic identity, 94.8% identified as White,

2.9% as Black or African American, and 1.0% as Asian. Five other racial or ethnic groups comprised a total of 1.3% of respondents. Average annual household income was $312,000, although, as with participant data, variability was high and many responses were missing

(SD=$310,000). General household respondents reported living in their present homes for an average of 17.1 years, a longer period than for participants, but only 14.8% reported previously paying for insecticidal treatment of landscape trees, compared with 56.6% for participants.

Respondents reported an average of 11.1 trees on their properties, between maintained and unmaintained areas, and spent an average of $463 annually on tree maintenance—values roughly half of participants.

Only 33 respondents overall (10.2%) had heard of the EABTP; the most commonly cited source of information was associates or friends (n=10), followed by TV or other media (n=9), and VDOF or VCE publications (n=7). Asked about the importance of preserving threatened landscape trees, respondents’ mean score was 4.0 on a scale from 1 (Extremely important) to 5

(Extremely important). When asked about the likelihood of regularly paying for treatment of threatened trees, respondents’ mean score was 3.8 on a scale 1 (Extremely unlikely) to 5

(Extremely likely). Lastly, when asked about interest in applying for cost-share assistance with landscape tree preservation, respondents’ mean score was also 3.8 on a scale from 1 (Not at all interested) to 5 (Extremely interested). These values are lower than those of parallel measure for participants, but are in regard to questions about a hypothetical pest threat. Table 7 summarizes respondent characteristics.

4.3.2 Tests for Nonresponse Bias

Tests of nonresponse bias by comparing means of external variables between respondent and nonrespondent groups. Two geographic, parcel-level variables and four demographic,

Census Block Group-level variables were compared for both participant and general household survey groups. For participants, no differences were evident between respondents and nonrespondents in any of the tests, indicating a final sample that was representative of all homeowner participants. For the survey of general households, tests indicated that educational attainment was significantly higher among respondents than nonrespondents, and that there were marginally significant differences in household income and proportion of minorities. Taken together, these results indicate that general household survey respondents formed a group that was more educated, and likely less wealthy and racially diverse than the sampling frame itself.

Complete results for tests of nonresponse bias are available in Appendix B.

Table 7. Summary of selected survey responses and external data for program participants and general households. EABTP participants General households Selected responses (n = 53) (n = 321) x̅ SD x̅ SD Located in Census Urban Area (%) 32.1 -- 100 -- Parcel size (acres) 30.624 80.646 0.442 0.574 Parcel tree cover (%) 58.6 26.8 45.0 15.8 Years since local EAB detection 2.4 2.2 4.7 4.4 Gender (% female) 39.6 -- 46.4 -- Age 58.8 12.0 58.7 12.0 Race or ethnicity (%) White 100 -- 94.8 -- Black or African American -- -- 2.9 -- Asian -- -- 1.0 -- Other ethnic or racial identity -- -- 1.0 -- Years of education 19 2.5 17.5 2.7 Annual household income ($) 232,000 160,000 312,000 310,000 Familiar with EAB (%) 100 -- 42.1 -- Aware of EABTP 100 -- 10.0 -- Years living at property 11.3 9.7 17.1 9.1 Number of trees on property 10.2 3.3 7.7 4 (maintained areas) Number of trees on property 8.3 5.6 3.4 5 (unmaintained areas) Annual tree maintenance budget ($) 987 961 463 923 Previously paid for tree pest 56.6 -- 14.8 -- treatment (%) Importance of preserving ash 4.4 0.7 4.0 0.9 trees/shade trees1 (1-5) Likelihood of regularly paying for 4.6 0.7 3.8 0.1 landscape tree treatment (1-5) Likelihood of applying for cost-share 4.6 0.8 3.8 0.1 funding (1-5) 1Survey language differed between respondent groups: program participants were asked about 'ash tree preservation' while general households were asked about 'shade tree preservation.'

4.3.3 Analysis of Response Mode

As a descriptive analysis, logistic regression was used to model the relationship of response mode (web or mail) as a function of the self-reported demographic characteristics of general household survey respondents. Variables entered as predictors in the model were Age,

Educational attainment, Household income, Gender, and Minority status. Assumptions of linearity of continuous variables were met, and no outliers were detected. Missing cases were deleted listwise, leaving only 231 cases in the analysis. The model was not statistically significant, χ2 (5) = 9.359, p = .096, correctly classifying only 60.2% of cases and explaining only 5.30% of the variance (Nagelkerke R2). However, Age was a statistically significant predictor (p = 0.049), negatively associated with the probability of a web response. This result suggests that younger respondents were more likely to complete the web rather than mail version of the survey, although the effect size was very small (Odds Tatio = 0.997) (Table 8). It is important to note, however, that this result provides no information regarding the effect of a mixed-mode survey on response rate. Logistic regression was conducted in SPSS.

Table 8. Binomial logistic regression results of mail vs. web responses to the general household survey using demographic predictors. Web responses were coded as 1 (n = 231). Independent Variables B S.E. p Odds Ratio Age -0.023 0.012 0.049 0.997 Years of education 0.027 0.053 0.610 1.028 Household income 0 0 0.310 1.000 Female (coded 1) -0.238 0.281 0.396 0.788 Minority status (coded 1) -0.627 0.555 0.258 0.534 2 Nagelkerke R 0.053 χ2 9.359 0.096 Bolded item indicates significance at the α = .05 level.

4.3.4 Factor Analysis of Attitudes Towards Urban Trees

An exploratory factor analysis of five, equally-scaled statements regarding the importance of urban trees yielded two factors. The first factor, termed Tree curb appeal, combined high loadings for attitudinal statements regarding the importance of urban trees to neighborhood character and property value (Cronbach’s α = 0.79). The second factor, termed

Tree affinity, combined high loadings for statements regarding benefits provided by trees, relative importance of trees in the private landscape, and a preference for preservation of old trees (Cronbach’s α = 0.67). Factor indices were then calculated casewise as the mean value of combined attitudinal measurements. These indices were employed as predictor variables in models of behavioral intention described below. Attitudinal statements, factor indices, and factor loadings are summarized in Appendix B.

4.3.5 Cluster Analysis of Ranked Motivations for Tree Preservation

Using a clustering technique for ranked-choice data, a 2-cluster solution for participants and 5-cluster solution for general households were selected. Clusters were formed around modal response patterns, and the relative distance of individual ordered responses from these patterns.

For participants, the order of responses was nearly identical: those in Cluster 1 indicated that the provision of shade was the most important reason for preserving threatened trees, while those in

Cluster 2 indicated that trees’ value an attractive part of the landscape is most important. Modal patterns for the five clusters selected for general households showed a greater diversity of opinion: high-ranked motivations included shade and landscape value, but also the provision of wildlife habitat, property value, and the prevention of species endangerment. Table 9 displays proportional membership, mean probability of membership, and modal ranked order for each cluster, along with relevant group characteristics. Cluster membership served as an additional categorical predictor of motivation, used in models of behavioral intention described below.

Table 9. Cluster analysis of ranked statements about motivations for tree preservation. Modal order of ranked moviations1 indicates most common ranking pattern of cluster members. Demographic mean values are included for comparison, and significant differences between clusters are noted. Program Participants General Households Cluster 1 Cluster 2 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Modal order of ranked 1,5,2, 5,1,2, 1,2,5, 1,5,3, 5,1,2, 2,5,1, 4,2,1, motivations 3,4,6 3,4,6 4,3,6 2,4,6 3,4,6 3,4,6 3,5,6

Number of respondents 16 37 117 33 113 51 6 Proportion of survey 0.293 0.707 0.366 0.104 0.353 0.158 0.018 respondents Mean age (binned 1-6) 4.538 4.974 4.813 4.516 5.038 4.755 4.75 Mean educational attainment 5.643 5.789 5.13 4.906 5.121 4.958 3.875* (binned 1-7) Mean household income 7.125 6.92 6.56 6.583 6.532 6.425 6 (binned 1-8) Mean number of maintained 2.143 2.692* 1.975 2.182 1.916 1.896 2.625* trees (binned 0-3) Mean number of unmaintained 1.643 2.205 0.653 1.242* 0.991* 0.75 1.5 trees (binned 0-3) Percent tree cover 0.589 0.584 0.426 0.451 0.461 0.479* 0.472 *Significantly different from the reference level of Cluster 1, within either survey group, at the α < .05 level. 1Motivations were given as: 1) Ash/Trees provide shade, 2) Ash/Trees provide wildlife habitat, 3) Ash/Trees can increase property values, 4) Ash/Trees might become rare or endangered, 5) Ash/Trees are an attractive part of landscaping, 6) Other, please specify. Program participants were asked specifically about ash trees while general households were asked about 'threatened landscape trees.’

4.3.6 Linear Models of Behavioral Intention

Assumptions of multiple linear regression were evaluated for all top-ranking models.

Independence of observations was verified using the Durbin-Watson statistic. Linear relationships were established between each dependent variable, full models, and independent variable, as was homoscedasticity of residuals. Tests for collinearity using calculated tolerance values did identify two pairs of highly correlated variables (Tolerance > 10) among participant data; high tolerance was resolved by dropping one variable from each pair. For all models, normality of residuals was visually confirmed from P-P plots. Tests did identify outliers, high leverage points, and influential points in all models. After an examination of the datasets, none of these unusual points appeared to be the result of error in measurement or data recording and were not removed.

Tree preservation intention

The objective of survey items asking about willingness to pay for regular landscape tree treatment was to measure the extent to which homeowners were prepared to take on tree health care as a recurring expense. For participants, the question addressed a present threat (EAB) to a specific tree genus; for general households the threat proposed was hypothetical, and no specific trees were named. Consequently, while each measure provides a reference point to the other, a statistical comparison between the two is not meaningful.

For participants, the top-ranked candidate model was Attitudes, comprised of seven individual predictors. The model itself was significant, although fit was relatively low (Adj. R2 =

0.137, p = .025). Within the model, there were two marginally significant predictors: first, a positive association with Importance of preserving ash trees (훽=0.275, p=.059), and secondly, a negative association with Factor A - Tree curb appeal (훽=-0.409, p=.059).

For general households, the top-ranked candidate model was the paired combination of

Attitudes + Property Characteristics. This model was significant and showed substantial goodness-of-fit (Adj. R2 = .412, p < .001). Here the second-ranking model, Attitudes, was considered a competitor since it ranked within 2 AICc of the first and was made up of fewer parameters. Significant predictors within the top-ranking model included a positive association with Importance of preserving landscape trees (훽=0.494, p <. 001) and negative association with the EAB infestation stratum Established (훽=0.129, p = .041). Comparisons of candidate models and summaries of top-ranked models are displayed in Tables 10-13.

Table 10. Program participants: Ranking of candidate linear models for the dependent variable Tree preservation intention. Number of respondents varied between models from 33 to 53 1 2 2 3 Model K Adj. R RMSE ΔAICc wi Attitudes 4 0.137 0.634 0 1 Attitudes + Property 11 0.095 0.649 16.749 0 Property Characteristics 7 -0.088 0.711 18.663 0 Personal Characteristics 6 0.079 0.655 24.979 0 Attitudes + Personal 10 0.043 0.667 35.392 0 Personal + Property 13 -0.091 0.712 49.81 0 Attitudes + Personal + Property 17 -0.19 0.744 74.21 0 Bolded text indicates top-ranking model. 1Number of model parameters. 2Second-order AIC, a small sample-corrected version of AIC, a measure of model fit which minimizes information loss. 3Akaike’s weight, a proportional measure of relative likelihood.

Table 11. Program participants: Summary of top-ranking linear model Attitudes for the dependent variable Tree preservation intention (n=53) Independent variables 훽 coeff. p Urban tree attitudes Factor A - Tree curb appeal -0.409 0.057 Urban tree attitudes Factor B - Tree affinity 0.413 0.071 Importance of preserving ash trees 0.275 0.059 Tree preservation motivation - Cluster 1 -- -- Tree preservation motivation - Cluster 2 0.16 0.224 Adjusted 푅2 0.137 F 3.059 0.025 Bolded item indicates significance at the α = .05 level.

Table 12. General households: Ranking of candidate linear models for the dependent variable Tree preservation intention. Number of respondents varied between models from 239 to 309. 1 2 2 3 Model K Adj. R RMSE ΔAICc wi Attitudes + Property 14 0.412 0.849 0 0.56 Attitudes 7 0.393 0.863 0.496 0.437 Attitudes + Personal 14 0.447 0.824 10.99 0.002 Attitudes + Personal + Property 21 0.443 0.827 21.254 0 Personal Characteristics 7 0.158 1.016 97.555 0 Personal + Property 14 0.143 1.026 109.621 0 Property Characteristics 7 0.077 1.064 127.356 0 Bolded text indicates top-ranking model. 1Number of model parameters. 2Second-order AIC, a small sample-corrected version of AIC, a measure of model fit which minimizes information loss. 3Akaike’s weight, a proportional measure of relative likelihood.

Table 13. General households: Summary of top-ranking linear model Attitudes + Property Characteristics for the dependent variable Tree preservation intention (n=309) Independent Variables 훽 coeff. p Attitudes Factor A - Tree curb appeal 0.109 0.059 Factor B - Tree affinity 0.092 0.135 Importance of preserving landscape trees 0.494 <.001 Tree preservation motivation - Cluster 1 -- -- Tree preservation motivation - Cluster 2 0.063 0.195 Tree preservation motivation - Cluster 3 0.023 0.652 Tree preservation motivation - Cluster 4 -0.025 0.603 Tree preservation motivation - Cluster 5 0.028 0.546 Property Age of house -0.066 0.189 Charac- Number of trees in maintained areas -0.015 0.744 teristics Number of trees in unmaintained areas -0.025 0.64 Parcel area -0.045 0.357 Parcel tree cover -0.001 0.983 EAB infestation strata - Undetected -- -- EAB infestation strata - Recent -0.062 0.271 EAB infestation strata - Established 0.129 0.041 Adjusted 푅2 0.412 F 15.963 <.001 Bolded items indicate significance at the α=.05 level.

Cost-share participation

Survey items asking about respondents’ level of interest in applying (or re-applying) for cost- share assistance were intended to measure the attractiveness of a public subsidy for preservation of personal landscape trees. Participants were asked directly whether they would re-apply for

EABTP in future years, while general households were asked about their interest in a hypothetical cost-share program, if trees on their properties were threatened by a future pest outbreak. As with models of Tree preservation intention, a comparison of predictive models for

Cost-share participation between participants and general households may be useful for reference, but statistical comparisons between the two are not meaningful.

Despite the differing scenarios presented to each respondent group, top-ranked candidate models were very similar. In both cases, the top-ranked model was predictor class of Attitudes, composed of seven individual predictor variables. For participants, the model was significant and fit moderately well (Adj. R2 = 0.223, p =.003). It contained a single significant parameter:

Importance of preserving ash trees (훽=0.513, p<.001). For general households, the equivalent model was significant and showed substantial goodness-of-fit (Adj. R2 = 0.456, p < .001). This model contained two significant, positively associated significant predictors: Importance of preserving landscape trees (훽=0.609, p<.001) and the factor index Tree curb appeal

(훽=0.110, p=.044). Comparisons of candidate models and summaries of top-ranked models for

Cost-share participation are available in Appendix B.

4.4 Discussion

Linear modeling of homeowner behavioral intention demonstrated the importance of measured attitudes in predicting homeowner intentions towards tree preservation and cost-share enrollment. In only one case did a predictor class other than Attitudes improve overall model fit.

This exception was seen in modeling the dependent variable Tree preservation intention among general households, where the EAB stratum Established was positively correlated with an intention to treat threatened trees. This finding suggests that the experience of living through a years-long EAB infestation may influence the behavior of some homeowners towards proactive tree preservation activity, regardless of whether they owned ash trees.

In most other cases where Attitudes alone proved the top-ranked model, the lone significant parameter within the model was the variable Importance of preserving ash trees or its equivalent regarding landscape trees. The single exception to this pattern was the addition of the factor index Tree curb appeal in a model of general household interest in applying to a hypothetical cost-share program. The association of this index with interest in cost-share funding suggests the importance of the neighborhood amenity value of trees as a motivating factor for homeowners seeking assistance for tree preservation.

However, in every top-ranked model the only parameter with a consistent, positive association with tree preservation and program participation was the attitudinal variable

Importance of preserving ash/shade trees. This variable originated from survey items which asked for 5-point ordinal response to questions regarding the abstract importance of ash tree (or shade tree) preservation in Virginia. Responses to these attitudinal questions proved to be the strongest and most consistent predictors of homeowners’ behavioral intentions, either to treat trees at their own expense or to apply for cost-share funding.

A peripheral topic of this research was an analysis of demographic characteristics by web or mail survey response mode. Logistic regression results indicated that the probability of a web response was negatively associated with age, although the effect size was small. This finding weakly supports the premise that younger respondents, when offered a choice between response modes, are more likely to complete a survey online. It does not, however, offer evidence to support the argument that offering a choice between response modes can boost survey response rates overall. A broader discussion of this topic and other study findings, implications, and directions for further study are included in Chapter 6.

CHAPTER 5 –PERCEPTIONS OF COST-SHARE PARTICIPATION AMONG FOREST PRACTITIONERS

5.1 Introduction

Conservation and natural resource incentive programs rely on the expertise of foresters and other resource professionals to meet with landowners and discuss land management objectives. In many cases, landowners report that the opportunity to talk with foresters in person and “walk the land” is a more valuable benefit of cost-share participation than financial assistance itself (Kilgore et al. 2007). Similarly, an evaluation of the Forestry Stewardship

Program in West Virginia found that the most common complaint among participating landowners was that foresters did not have time to visit their properties often enough (Egan et al.

2001). For these reasons, I aimed to broaden an analysis of Virginia Department of Forestry’s

(VDOF) 2018 Emerald Ash Borer Treatment Program (EABTP) by examining the role of forest practitioners in program implementation.

State- and county-level foresters have a decades-long history of providing technical assistance and administering forestry incentives among landowners (Esseks and Moulton 2000), although almost exclusively in rural areas. Conversely, commercial arborists work primarily for urban property owners, and for many, a majority of business revenue is derived from management of tree pests and pathogens. Since Virginia’s EABTP provided for funding of tree treatment in any location, urban or rural, the program made use of the expertise of both practitioner groups.

Arborists were involved at two stages of the EABTP: first in submitting a bid, then later conducting the treatment. Program policies required that homeowners could only apply for funding after receiving a signed bid for treatment from a licensed pesticide applicator. Of 107 approved applications, only 3 involved landscape contractors who were not arborists (Bean

2019). Once an application was submitted, a homeowner could only schedule tree treatment after receiving on-site approval from a VDOF forester. For some arborists, months may have elapsed between bidding and conducting treatment because of VDOF stipulations regarding the timing of treatment (VDOF 2018). After conducting tree treatments, arborists were paid in full by the homeowner, as for any other contracted work. VDOF’s reimbursement payment to a homeowner was disbursed only after the homeowner submitted a zero-balance invoice from the arborist.

County foresters were tasked with on-site approval of tree treatment requested by homeowners and bid on by arborists. Foresters’ responsibilities for each tree proposed for treatment included (1) verifying tree genus and species, (2) confirming a minimum trunk diameter of twelve inches, and (3) assessing whether tree vitality was high enough to warrant treatment (VDOF 2018d). The threshold of tree vitality adopted by the EABTP a maximum of

30% crown dieback, a rule of thumb recommended by researchers (Bick et al. 2018). To ensure reliability of assessment, VDOF held training sessions prior to program launch with instruction and a photo guide of incremental ash tree canopy loss (Chamberlin 2018a). Of 107 total site assessments, 17 were conducted by foresters from the Forest Health division at VDOF headquarters. The remaining 90 assessments were conducted by 24 county foresters, almost all of whom held the positions of Area Forester or Senior Area Forester.

Participation in the EABTP did not dramatically alter job descriptions of either arborists or foresters but did temporarily change the day-to-day context for both. Foresters were asked to assess properties a few acres in size instead of a few hundred acres, and to scrutinize the work of tree services instead of logging contractors. For arborists, EABTP participation required waiting weeks to months for VDOF approval but offered the possibility of greater demand for ash tree injection.

As intermediaries between program administrators and participants, practitioners had great influence on the program’s initial year. This is evidenced by the high proportions of program participants who cited arborists (25%) and VDOF or VCE employees (23%) as their initial source of EABTP information. Consequently, practitioners’ engagement with the program will continue to affect its future success. The objective of this research was to inform future initiatives by examining the engagement of forest practitioners with the EABTP, and in urban forest health more broadly. Topics investigated included potential associations between professional experience and perceptions of threats posed by forest pests; predictors of interest in future program participation, and specific reasons for interest or disinterest in participation.

5.2 Methods

5.2.1 Construction of Survey Sampling Frames

Survey research was conducted with samples drawn from both VDOF county foresters and Virginia arborists. Unlike homeowner surveys, in which recipients were contacted directly by mail and email, county foresters and arborists were contacted with survey requests through coordinating agencies. This method provided relatively large contact lists, but less information about recipient characteristics, compared to homeowner sampling frames.

The forester sampling frame included all VDOF county foresters who carried responsibility in reviewing EABTP applications, defined as those holding the positions of Area

Forester or Senior Area Forester. This group was estimated to include 105 foresters (Bean 2019).

On behalf of this research project, VDOF Forest Health officials contacted all members of the sampling frame with survey requests, effectively conducting a complete sample. From this sample, a total of 19 usable responses were recorded, for a response rate of 18.1%. For a 50/50 split response at 95% confidence level, these figures correspond to a 20.4% margin of error,

meaning that survey results can only be generalized to the sampling frame with a large amount of uncertainty (±20.4%).

Similarly, arborists were contacted through the coordinating agency of MAC-ISA, the

Mid-Atlantic chapter of the International Society of Arboriculture. While the term arborist generically refers to a landscape tree care professional, Certified Arborist refers to those certified through the ISA. the sampling frame was defined less stringently to include all arborists (tree care professionals) with a business address or clients in Virginia. There were 1081 Certified

Arborists in Virginia in a MAC-ISA database who formed the primary contact list, and who were contacted by email with survey requests. Follow-up survey recruitment posts in a MAC-ISA online newsletter expanded the contact list to include all MAC-ISA members (n=1139), whose membership overlapped with the primary contact list of Virginia Certified Arborists, and also included non-certified professionals, and members outside of Virginia. After accounting for recipients on both lists, the total number of unique recipients was calculated to be 1590. A total of 144 usable responses were recorded, for a response rate of 9.1%. Two specific survey items allowed for post-hoc parsing of response data by state location and certification status. Of the

144 respondents, 134 reported either having a business or clients located in Virginia. For a 50/50 split response at a 95% confidence level, these figures correspond to an 7.8% margin of error.

5.2.2 Construction of Survey Instruments

Survey instruments for foresters and arborists were developed using Qualtrics and shared most survey items. The foresters’ survey contained 22 questions and had an estimated response time of 12 minutes, while the arborists’ survey contained 30 questions, and had an estimated response time of 13 minutes. As with homeowner surveys, forest practitioner surveys were designed to assess respondents’ experience with EAB, attitudes towards urban trees, and

intentions regarding future EABTP participation. A limited number of demographic characteristics were requested, along with professional characteristics such as certification status, years of experience, and job title. Finally, a series of ranking questions were included that measured respondents’ relative preference for motivations for ash tree preservation, landscape tree species, perceptions of forest pest threats, and on-the-job recommendations for EAB management. A summary of survey items and measurement scales in included in Appendix C, while the full survey instrument is available in Appendix G.

Because forester and arborist survey instruments shared most questions and wording, a single pre-test instrument was sent by email to a group of twenty-five Certified Arborists located within and outside of Virginia. Response data and comments from fifteen responses informed editing of final survey versions, including improvements in clarity of wording and logical flow.

5.2.3 Data Collection

For both surveys, recipients were initially contacted with survey requests beginning in late September 2018, and with follow-up contacts through the end of October. All survey recruitment materials and instruments were submitted first to the Virginia Tech IRB and subsequently approved by Western IRB. An IRB Exemption letter is included in Appendix H.

Foresters who had previously participated in the program (n = 25) were first contacted by email in late September 2018, with a survey request message and unique URL sent by VDOF on behalf of this project. A follow-up survey recruitment post and link were included in an online

VDOF newsletter on October 2, sent to all VDOF employees (n ≈ 230). While the total number of newsletter recipients was greater than the size of intended sampling frame (n = 105), the post

explained that only foresters with potential EABTP responsibility were encouraged to take the survey.

Four separate survey requests were sent to arborists by MAC-ISA on behalf of this project, beginning with a recruitment post in an online newsletter to chapter members in late

September 2018. Following this, two survey requests were emailed to Certified Arborists located in Virginia, spaced by one-week intervals. Finally, a recruitment post included in a second online newsletter was sent to all chapter members in late October. After approximately three months, survey response collection for both surveys was closed on December 31, 2018.

5.2.4 Data Analysis

Tests for Nonresponse Bias

To create a reference group of foresters against which to compare survey respondents, data were compiled from VDOF online records for all county foresters with the position of

Senior Area Forester or Area Forester—positions with responsibility for EABTP application review. This search yielded 64 foresters. Frequencies of foresters’ Job title, Gender, and

Location by VDOF work area were recorded, and tested against the same characteristics for forester survey respondents using a chi-square test of independence.

Testing for nonresponse bias in the arborists’ survey response was complicated by the fact that two contact lists with overlapping membership were used to distribute recruitment materials. However, with anonymized membership data provided by MAC-ISA, limited tests for nonresponse bias were conducted, again using chi-square tests of independence between respondents within Virginia and a reference group of Certified Arborists within Virginia.

Frequencies were compared by group for professional Area of practice and for county location, aggregated by EAB strata. All chi-square tests were conducted in SPSS.

Factor Analysis of Attitudes Towards Urban Trees

As with analysis of homeowner response data (Chapter 3), an Exploratory Factor

Analysis was conducted of five equally-scaled survey items regarding attitudes towards urban trees. For this analysis, responses were aggregated between arborists and foresters to facilitate cross-sectional comparison. From a scree plot of eigenvalues, a 2-factor solution was selected.

Analysis was conducted in R, using the packages ‘psych,’ and ‘GPArotation.’

Cluster Analyses: Motivations, Recommendations, and Perceptions of Pest Risk

Cluster analysis was employed to partition respondents according to three ranked-choice survey items. Analyses were conducted using aggregated forester and arborist data, for survey items regarding motivations for preserving ash trees, and for perceptions of threats posed by specific invasive forest pests. For arborist response data only, a third cluster analysis was conducted of ordered response data regarding EAB management recommendations.

For each cluster analysis, response data collected consisted ordered vectors of numbers representing as many as seven ranked options. Using R package ‘Rankcluster’ the optimal number of clusters was selected for each analysis from a plot of BIC against an increasing number of clusters. Cluster membership was assigned to individual cases and treated as nominal variables in further analyses.

Analysis of Pest Risk Perceptions

Understanding differing perceptions of risks posed by pests in relation to specific professional characteristics might be relevant to recruiting practitioner support for future iniatives. Many reasons could account for diverging perceptions of risk—for instance, in pockets of Virginia that remain unaffected by EAB, perceptions of the severity of EAB damage may be lower. Alternately, the recent arrival of ‘new’ pests, might alter the prioritization of resources for some foresters. One relevant example is the spotted lanternfly (Lycorma deliculata White) first detected in 2018 in Virginia (Day et al. 2018). Similarly, it is possible that experienced foresters and arborists may have a better understanding of risks posed by longstanding pest infestations such as the gypsy moth, which may currently attract less research attention and funding than recent arrivals.

Using chi-square analyses, null hypotheses of no association were tested between perceptions of pest risk and five variables representing professional experience. Perceptions of pest risk were represented by pest threat cluster membership referenced above, forming a a four- level categorical variable Pest threat cluster. Five variables representing aspects of professional experience, recoded in categorical form, were tested against these clusters for association. These included EAB strata, representing location within Virginia by local age of EAB infestation, Work experience, representing years of professional experience, Clientele acreage, representing the percentage of clients with properties larger than five acres, EAB experience, representing frequency of prior EAB-related work, and Practitioner type, representing a respondent’s status as a forester or arborist. As a series of multiple comparisons with the same dependent variable, a

Bonferroni correction was necessary to reduce family-wise Type I error, lowering the significance level to α =0.01. Chi-square tests were conducted in SPSS.

Linear Models of Program Participation

While survey response data suggested that foresters expressed greater interest in cost- share participation that arborists, this comparison is not necessarily meaningful. First, the large difference in sample size complicates a comparison of means. Secondly, interest in program participation expressed by foresters, whose job duties may have required participation, were not equivalent to responses from arborists, who had no obligation to the program. For these reasons, only arborist response data were included in models of program participation. Additionally, respondents without a business address or clients in Virginia were excluded from analysis, resulting in a sample size of 134.

Using data drawn from survey responses and external sources, candidate models were fitted to the dependent variable of Interest in program participation. This variable was drawn from a single, ordinal survey item which asked arborists about their interest in future EABTP participation, either in pricing work or carrying out insecticidal applications (see Appendix C for summary of survey items or Appendix G for full survey instrument).

As with models of homeowner intentions, candidate models were constructed from groups of three to eight individual predictors categorized by predictor class: Professional

Characteristics, Personal Characteristics, and Attitudes. Seven candidate models were constructed from the combination of individual, paired, and grouped predictor classes. Each of these models was fitted to the dependent variable using hierarchical multiple linear regression in

SPSS. Model selection was accomplished by ranking of candidate models in order of decreasing

Akaike’s weight (wi), a proportional measure of relative likelihood derived from AIC.

Qualitative Analysis of Written Responses

In addition to rating their interest in program participation, practitioners were also asked to state their reasons for interest, or lack of interest in EABTP participation. Written responses provided additional insight into practitioners’ level of engagement with EABTP. Of 163 aggregate respondents to both practitioners’ surveys, 98 entered a response for this item. These responses were coded by twelve common themes, and graded as positive, neutral, or negative with regard to interest in program participation.

5.3 Results

5.3.1 Summary of Response Data

Respondents to the foresters’ survey (n=19) were predominantly male (78%), in their mid-forties (x̅ = 44.0 years), and had completed a graduate degree (x̅ = 17.6 years of education).

Responding foresters represented 15 different counties, where on average, EAB had been detected recently (x̅ = 4.2 years since detection). Respondents most commonly held the position of Area Forester and reported an average of 18.5 years of experience in forestry. While only

10.5% held Society of American Foresters (SAF) certification, 73.7% held ISA certification.

Regarding their work prior to 2018, foresters reported that on average, 20.0% of their working hours were dedicated to forest health-related projects and indicated their frequency of EAB- related work as 2.5, on a scale from 1 (Very rarely) to 5 (Very frequently). Additionally, foresters reported that on average, 78.2% of landowners they interacted with on the job owned more than five acres.

All respondents reported making EABTP-related property visits to review applications.

The number of properties visited ranged one to twenty, averaging between five and six. Forester

respondents’ average level of interest in future program participation was 3.7, on a scale of 1

(Not at all interested) to 5 (Extremely interested). Respondents’ location by county are displayed in Figure 10; selected survey response data are summarized in Table 14. Additional attitudinal response data are discussed below in relation to factor and cluster analyses.

Most arborist respondents (93.1%) reported that either their business or clients were located in Virginia. The remaining 10 (6.9%) out-of-state responses were retained for initial analyses but excluded from models of EABTP participation. Respondents were predominantly male (83%), on average in their late forties (x̅ = 47.2 years) and had continued their education beyond a four-year degree (x̅ = 16.9 years of education). Arborist respondents represented 49 different Virginia counties, where on average, the EAB infestation was established (x̅ = 6.8 years since detection). Respondents most commonly described their area of practice as Urban forestry/Government and had on average 18.9 years of arboricultural experience. As with forester respondents, a low percentage (7.0%) held SAF certification, but almost all (98.6%) held ISA certification.

Regarding their work prior to 2018, arborists reported that on average, 33.0% of working hours were dedicated to forest health, and indicated the frequency of EAB-related work as 3.3, on a scale from 1 (very rarely) to 5 (very frequently). Arborists also reported, on average, that

35.7% of landowners they interacted with on the job owned more than five acres. Of 144 respondents, 76 (52.3%) reported previously being aware of EABTP—of these, the primary source of information was VDOF or VCE employees (36.8%), followed by associates or friends

(34.2%), and VDOF or VCE publications (26.3). Only nine respondents (6.3%) reported making

EABTP-related property visits, either to submit bids or treat trees. The number of properties visited ranged from one to three. Arborist respondents’ mean level of interest in future program

participation was 2.8, on a scale of 1 (Not at all interested) to 5 (Extremely interested), lower than the equivalent response for foresters. County locations of Virginia arborist respondents are displayed in Figure 10; selected survey responses are summarized in Table 14.

Table 14. Summary of selected survey responses and external data for foresters and arborists. Foresters Arborists Selected responses (n = 19) (n = 144) x̅ SD x̅ SD Years since local EAB detection1 4.2 3.7 6.8 3.9 Gender (% female) 22.2 -- 17.7 -- Age 44.0 11.8 47.2 12.7 Years of education 17.6 1.6 16.9 2.1 ISA Certified Arborist 73.7 -- 98.6 -- SAF Certified Forester 10.5 -- 7.0 -- Years of professional experience 18.5 12.0 18.9 12.0 Percentage of client properties > 5 acres 78.2 19.4 35.7 30.5 Percentage of work hours spent on forest health 20.0 14.8 33.0 26.1 Frequency of prior EAB-related work (1-5) 2.5 1.1 3.3 1.2 Previously aware of EABTP (%) -- -- 52.3 -- Number of EABTP site visits or bids 5.2 5.6 0.2 1.2 Interest in future EABTP participation (1-5) 3.7 0.8 2.8 1.3 1There were many survey responses with unknown locations; estimates based on n=11 for foresters and n=111 for arborists.

Figure 10. Map of Virginia displaying counties by EAB infestation strata (defined by year of initial EAB detection), and rough locations of forester (n=19) and arborist (n=134) survey respondents.

5.3.2 Tests for Nonresponse Bias

Tests for nonresponse bias using limited data available for both practitioner groups provided evidence of a representative response for foresters but raised questions about nonresponse bias among arborists. Between forester respondents and a forester reference group, chi-square tests showed no associations between group and Gender, Job title, or Location. While the foresters’ survey drew only 19 responses, this number represents 76% of all VDOF county foresters who participated in the 2018 EABTP. From this limited analysis, survey responses appear to represent the gender ratio, experience level, and geographic distribution of a broader group of ranking county foresters. Test results are summarized in Appendix C.

For arborist repondents, chi-square tests indicated significant associations between group membership and both variables tested (Area of practice and EAB strata), although with only moderate effect sizes, as measured by Cramer’s V. Post-hoc assessment of standardized residuals

indicated that frequencies of arborist respondents in listing their area of practice as

Research/Education/Training was higher than in the reference group, and lower than the reference group for the category Other. This category represented landscape contractors, landscape architects, and nursery owners—professionals who are less likely to be involved in treatment of tree pests. Finally, frequencies of arborist respondents in counties within the

Undetected EAB infestation strata were also lower than among the reference group. Taken together, these measures of association indicate that arborist survey respondents as a group, relative to the sampling frame of all Virginia MAC-ISA members, included fewer landscape professionals, a greater number of researchers or educators, and fewer people located in southeastern Virginia counties. Test results are summarized in Appendix C.

5.3.3 Factor Analysis of Attitudes Towards Urban Trees

Factor loadings of survey items on the two factors closely matched those of homeowner responses. The first factor, again termed Tree curb appeal, combined high loadings for responses to the belief statements regarding neighborhood character and property value (Cronbach’s α =

0.67). The second factor, termed Tree affinity combined high loadings for responses to belief statements regarding benefits provided by trees, relative importance of trees in the private landscape, and a preference for preservation of old trees (Cronbach’s α = 0.95). Factor indices were then calculated casewise as the mean value of the combined belief statements. Attitudinal statements, factors, associated loadings are summarized in Appendix C.

5.3.4 Cluster Analyses: Motivations, Recommendations, and Pest Risk

Cluster analysis of aggregated forester and arborist response data regarding ranked motivations for ash tree preservation yielded four clusters. Unlike ranked homeowner motivations, which focused on trees’ value as landscape elements, top-ranked motivations included ash trees’ provision of shade, and the necessity of preventing ash species from becoming rare. In only one of four clusters was ash trees’ contribution to property value ranked in the top three. A summary of cluster analysis of ranked ash preservation motivations is included in Appendix C.

Cluster analysis of ranked responses regarding arborists’ most frequent recommendations for EAB management yielded two clusters. Ranked recommendation patterns were almost identical: the insecticidal options of imadicloprid and emamectin were ranked at the bottom for both clusters, while the two top places were held by the options of ‘wait and see’ or tree removal.

A summary of cluster analysis of arborists’ EAB management recommendations is included in

Appendix C.

Finally, cluster analysis of aggregated forester and arborist data regarding their ranked perceptions of threats posed by seven invasive forest pests yielded four clusters. In three out of four clusters, EAB was ranked as the top threat, followed either by hemlock woolly adelgid or

Asian longhorned beetle. The remaining cluster, which was also the largest single grouping, ranked the southern pine beetle as the most threatening pest, followed by the spotted lanternfly.

Table 15 summarizes cluster analysis of aggregated pest threat rankings.

Table 15. Aggregated forester and arborist response data: Cluster analysis of forest pests ranked by level of perceived threat to Virginia's forests. Modal patterns represent most common ranking of each cluster. Cluster 1 Cluster 2 Cluster 3 Cluster 4 Modal pattern of ranked pests1 6,1,2,3,4,5,7 6,1,4,3,5,7,2 5,4,1,2,3,6,7 6,2,4,3,7,1,5 Number of respondents 53 30 70 10 Proportion of survey 0.324 0.184 0.429 0.063 respondents 1Forest pests were given as: 1) hemlock woolly adelgid, 2) Asian longhorned beetle, 3) gypsy moth, 4) spotted lanternfly, 5) southern pine beetle, 6) emerald ash borer, 7) walnut twig beetle

5.3.5 Chi-square Analysis of Pest Risk Perceptions

Minimal assumptions for chi-square analysis were met for all tests: in each crosstabulation, 80% of cells contained expected values greater than 5, and there were no expected frequencies less than 1. Tests indicated no associations between any of the professional experience variables and perceptions of pest threats, at the α = 0.01 level of significance. The only variable approaching a significant association and moderate effect size was Clientele acreage (p = 0.077, V = 0.200). This variable was originally recorded as a continuous measure, representing the percentage of clientele who owned more than five acres, and was recoded into three bins (0-33%, 34-66,% 67-100%) for this analysis. Test results from chi-square analysis of pest threat perceptions are displayed in Table 16.

Table 16. Chi-square analyses of independence between practitioner pest threat clusters and professional characteristics. Variables listed represent measured aspects of professional experience, recoded in categorical form. Four pest threat clusters were tested against each variable listed for independence. Number of cases varied from 142 to 163.

Professional characteristics Pearson's χ2 df p* Cramer's V

EAB strata 4.505 6 0.609 0.124

Work experience 8.158 6 0.227 0.158

Clientele acreage 13.382 6 0.077 0.200

EAB experience 11.566 12 0.481 0.161

Practitioner type 2.400 3 0.494 0.121 *Tested against a Bonferroni-corrected significance level of α = 0.01

5.3.6 Linear Models of Program Participation

The data met multiple linear regression assumptions for independence of observations, linearity, homoscedasticity, lack of collinear predictors, and normality. No outliers or influential points were identified; however, 11 points with high leverage were identified. Since a review of the data provided no evidence of error, no cases were removed.

Of seven candidate models for the dependent variable Interest in program participation, the top-ranked model by Akaike’s weight (wi) was Professional characteristics, although this model was poorly-fitting and not significant (Adj. R2 = 0.033, p = 0.166)). An equivalent model within 2 AICc of the top-ranked model was Attitudes, also with poor fit and nonsignificant F- ratio (Adj. R2 = 0.034, p = 0.139) Table 17 displays model fit, error, and ΔAICc for all candidate models.

Within the model Professional characteristics, the lone significant predictor, EAB management recommendation - Cluster 2 (훽=0.627, p = 0.02) was positively correlated with program participation. This binomial variable grouped respondents into two clusters according to

ranked preference for EAB management strategies—those in Cluster 2 preferred a “wait-and- see’ approach over tree removal. Table 18 displays a summary of the model Professional characteristics. There were no significant predictors with the equivalently-ranked model

Attitudes, although the factor index Tree affinity was marginally significant (훽=0.222, p =.055).

This multi-item index combined agreement with statements regarding benefits provided by trees, their relative importance next to other landscape elements, and a preference for preservation of declining trees. A second, marginally significant predictor within this model was Sales expectations (훽=0.157, p =.099). The positive correlation here indicates an expectation that participation might be good for business. Table 19 displays a summary of the model Attitudes.

Table 17. Arborist response data: candidate models for Interest in future program participation 2 Model K Adj. R RMSE ΔAICc wi Professional characteristics 8 0.033 1.269 0.000 0.422 Attitudes 5 0.034 1.269 0.578 0.316 Personal characteristics 3 -0.026 1.307 2.010 0.154 Personal + Attitudes 8 0.013 1.282 4.538 0.044 Professional + Attitudes 13 0.055 1.255 5.132 0.032 Personal + Professional 11 0.200 1.278 5.405 0.028 Personal + Professional + Attitudes 16 0.054 1.256 9.720 0.003 Bolded rows show top-ranking models. The model Attitudes is considered a competitor to Professional characteristics since it is within 2 AICc and has fewer parameters.

Table 18. Arborist response data: Summary of model Professional characteristics for the dependent variable Interest in future program participation. Independent variables 훽 coeff. p EAB infestation strata - Undetected -- -- EAB infestation strata - Recent -0.507 0.279 EAB infestation strata - Established -0.223 0.645 Years of professional experience -0.006 0.607 Percentage of clientele with < 5 acres -0.003 0.432 Percentage of forest health work hours 0.006 0.201 Frequency of prior EAB-related work 0.034 0.781 EAB management recommendation cluster 0.627 0.020 Adjusted R2 0.033 F 1.582 0.166 Bolded items indicate significance at the α=.05 level.

Table 19. Arborist response data: Summary of the model Attitudes for the dependent variable Interest in future program participation. Independent variables 훽 coeff. p Factor index 1 - Tree curb appeal -0.066 0.566 Factor index 2 - Tree affinity 0.222 0.055 Ash preservation motivations - Cluster 1 -- -- Ash preservation motivations - Cluster 2 0.063 0.571 Ash preservation motivations - Cluster 3 0.116 0.368 Ash preservation motivations - Cluster 4 -0.085 0.480 Expected effect of EABTP on sales 0.157 0.099 Adjusted R2 0.034 F 1.658 0.139

5.3.7 Qualitative Analysis of Written Responses

Aggregate forester and arborist written responses regarding level of interest in program participation were coded by 12 common themes, and each graded as positive (45), neutral (24), or negative (29). The two most common themes given for program interest were the benefits of

EABTP towards ash preservation (n=19), and the assistance the program provided for clients

(19). Some examples of reasons for interest in the program, included:

“We need to preserve a large enough population to allow development of genetic resistance…”

“Giving my clients more options for treatment is always a better way to go.”

“We take care of many trees. If our clients wish to participate, that is great.”

The two most common themes in reasons given for lack of interest were a regional low abundance of ash trees (n=12), and the perceived unsustainability of the program model (n=6).

Some examples of reasons for lack interest in the program included:

“Consider EAB a wakeup call to replace tree diversity.”

“More government paperwork to deal with. The treatments I perform work well.”

“Most of the ash that I see in Fairfax County are too far gone…”

Finally, written responses graded as neutral consistently expressed the idea that participation in the EABTP was not relevant to the respondent’s job duties. Many of these respondents were employees or utility contractors whose work would not take them into contact with residential clients.

5.4 Discussion

While EAB management was the top-ranked pest threat for most forest practitioners, it was one of several competing concerns. Newly detected pests, new discoveries in research, and changing ecological conditions all contribute to forest practitioners’ understanding of risk, which may differ from that of the public (Liebhold 2012). Tests for association between professional characteristics and perceptions of pest risk did not, however, identify any strong links between professional experience and perception of risk.

Models of arborists’ interest in future EABTP participation were poorly fitting and not statistically significant. Significant and marginally significant predictors identified within top- ranked models are of little substantive value, but perhaps suggest directions for future research.

The significance of the second EAB management recommendation cluster in a model of program participation may indicate that preference for a slower-moving approach to pest management is associated with interest in EABTP. Alternately, the inverse relationship may provide a clearer explanation: a preference for preemptive tree removal may be associated with low interest in

EABTP participation. Other marginally significant predictors of program participation included on one hand, Tree affinity, a factor index favoring a non-market valuation of urban trees, and on the other, Sales expectation, emphasizing the profitability of program participation.

Written responses contributed to a more nuanced understanding of practitioners’ attitudes towards EABTP. Of those who perceived the program as relevant to their work, positive responses outweighed negative responses by a ratio of about 3:2. The primary reasons given for interest in participation was the program’s importance for ash preservation and how it may benefit clients. Most practitioners who responded with reasons for lack of interest in EABTP participation did so based on low abundance of ash trees in their localities, whether due to historically low ash abundance or an established EAB infestation. A few, however, expressed a lack of interest in the EABTP based on a perception of its unsustainability—specifically, six respondents stated a belief in the economic or ecological inefficiency of EABTP funding. A broader discussion of this topic and other study findings, implications, and directions for further study are included in Chapter 6.

CHAPTER 6 – THESIS CONCLUSION

6.1 Summary of findings

6.1.1 Urban Participant Properties

Properties of urban EABTP participants (n=28) were relatively large and wooded, with a mean parcel size of 0.587 acres and tree cover of 48%, and on average, were developed in the mid-20th century (mean house age = 61.6 years). A cluster analysis of 16 inventoried properties using these three variables partitioned properties into three groups, termed Wooded exurban,

Historic urban, and Contemporary suburban. Both parcel size and tree cover were highest for the first cluster and lowest for the second.

A total of 365 trees and 57 species were inventoried across 16 properties visited. Mean species abundance was 10.1, with a mean Shannon index of H’ = 1.99. Across all sites, white ash

(Fraxinus americana) outperformed all other species in terms of relative abundance (16.6%), relative basal area (34.3%) and other related indices. By contrast, green ash (Fraxinus pennsylvanica) was barely present—represented by a single tree on one site. Measured by relative abundance, other species commonly found across sites were red maple (Acer rubrum L.–

8.5%), hackberry (Celtis occidentalis L. – 4.9%), flowering dogwood (Cornus florida – 4.4%), and silver maple (Acer saccharinum L.– 4.2%).

Comparisons of species composition between participant properties and surrounding urban forests indicated clear differences. Measured as a percent of total, ash relative abundance on urban participant properties was almost four times greater than in residential areas of surrounding municipalities. Measured by relative structural value, ash on participant properties also outvalued those in surrounding municipalities by a factor of three. Data for surrounding municipalities were drawn from urban forest assessments conducted in 2010-2011, a period

during which the ash borer was not yet present in most of Virginia. This fact serves to emphasize the unusually high number of ash trees on participant properties, relative even to pre-EAB levels.

Linear models of site species composition as a function of property characteristics were not successful in fitting the data. A model of ash Relative importance value fit the data poorly

(Adj. R2 = -0.095) and did not contain any significant predictors. A second model of ash Relative structural value showed slightly better fit (Adj. R2 = 0.185) and contained one marginally significant predictor, Years since local EAB detection (p = 0.093).

6.1.2 Homeowner Engagement in Landscape Tree Preservation

Survey research conducted with homeowner participants of the EABTP and a general household population indicated broad support for personal investment in tree preservation and broad interest in cost-share funding. Respondents to both surveys were predominantly white, majority male, and in their late fifties. Apparent differences between the two groups included the much larger average parcel size of primarily rural participant respondents (32.1 vs. 0.44 acres), and participants’ lower average household income ($232,000 vs. $312,000). Participants also reported, on average, two more years of education than general household respondents. An assessment of whether choice of survey response mode (mail or web) was influence by demographic characteristics did not demonstrate any strong effects, although a small association was evident between increasing age and a web response (p = .049, Odds Ratio = 0.977).

Models of behavioral intention consistently demonstrated the importance of attitudes in predicting behavior, as compared with personal or property characteristics. Significant or marginally-signficant predictors of willingness to pay for regular tree treatment included an abstract agreement with the importance of preserving trees for both survey groups. For

participants, this variable was joined by a negative association with the factor index Tree curb appeal, while for general households, location in the Established EAB strata was also positively associated. For models of interest in future cost-share enrollment, agreement with the abstract importance of tree preservation was again significant for both survey groups. Among general households, the factor index Tree curb appeal was also positively associated.

6.1.3 Practitioner Perceptions of Cost-Share Participation

Survey research conducted with VDOF county foresters and Virginia arborists demonstrated many similarities between two groups. Both groups reported an equivalent level of professional experience (x̅ =19 years), with a mean of 18 years of education for foresters, and 17 years for arborists. Responding foresters were also apparently younger (x̅ =44 years) than arborists (x̅ =47 years), and there was a greater proportion of female respondents among foresters

(22%) than arborists (17%). Few from either group held SAF certification (11% of foresters, 7% of arborists), while majorities held ISA certification (74 and 98%, respectively). Notably, while not directly comparable measures, interest in cost-share participation appeared higher among foresters (x̅ =3.7 out of 5) than arborists (x̅ =2.8).

A cluster analysis of practitioners’ ranked perceptions of threats posed by seven invasive pests grouped respondents in four clusters based on commonalities in ranking patterns.

Membership in these clusters was tested for association against a series of five professional characteristics, to assess the influence of these factors on pest management priorities. No significant associations were detected.

Models of arborist interest in program participation were developed using three predictor classes: Professional characteristics, Personal characteristics, and Attitudes. Neither of the two

top-ranking models were significant or fit the data well. However, within the model Attitudes,

Tree affinity was a significant predictor (p=0.055), potentially indicating the importance of non- market valuation of trees in program participation. Other marginally significant predictors included a preference for ‘wait and see’ EAB management approach, and the expectation of increased sales due to program participation.

Written responses regarded interest or lack of interest in the program demonstrated that a number of survey respondents did not work in a position where EABTP was directly relevant. Of those for whom the program was relevant, positive responses outweighed negative ones by a ratio of 3:2. The most common reason for interest of participation was to promote ash preservation; the most common reason for lack of interest was an insufficient local ash tree population. .

6.2 Implications

6.2.1 Urban Participant Properties

Comparisons of species composition between sites and cities demonstrated that ash trees were roughly three times more abundant and valuable on participant properties than in surrounding urban forests. This contrast suggests an important research topic—the identification of a threshold value or modeled relationship between site ash abundance and tree preservation actions. For urban forest managers, this information, combined with detailed urban forest inventory data could inform an efficient outreach strategy. For instance, in 2009 the city of

Milwaukee conducted a remote tree inventory of the entire urban forest using hyperspectral and

LiDAR data. With ash trees identified on private property, officials were able to alert close to

30,000 homeowners of the risk posed by EAB and the need to treat or remove trees (Sivyer

2010). Efforts such as these could be fine-tuned with information regarding which homeowners are most likely to act based on a threshold number or size of threatened trees.

Average characteristics of urban participant properties describe a relatively large, wooded, mid-century development. Notably, participants in urban areas made up less than one- third of all individual homeowner participants. From this we can gather that in the EABTP’s first year, funding was directed primarily to rural areas, and secondarily to large, wooded properties on urban peripheries. Cluster analysis further helped illustrate that among urban properties

(n=28), likely only one-third of these were in moderately dense neighborhoods (Cluster 2 median parcel size = 0.39 acres).

The geographic trends identified have positive implications for preservation of ash genetic diversity and some tree benefits. Compared to smaller, urban lots where most trees are likely to be clonally-reproduced cultivars, on rural and wooded exurban properties, many more trees have naturally regenerated from open-pollinated seed. The genetic diversity preserved in these trees represents an important resource for research (see Koch et al. 2015) and the ongoing survival of ash species. Additionally, the above-ground biomass of forest-grown trees is greater than of open-grown, maintained urban trees (Nowak et al. 2013). This implies that preservation of ash trees on wooded, urban edges will result in greater carbon storage than equivalent actions in more densely populated neighborhoods.

However, since an individual tree’s canopy is limited by neighboring trees and may even retreat in the presence of fast-growing ones (Spector and Putz 2006), net tree benefits are not necessarily maximized by funding preservation of trees on predominantly wooded lots. Since tree benefits such as energy savings, runoff avoidance, and pollution removal are related to the physical dimensions of a tree’s crown (Lee et al. 2016), trees in lower-canopy areas may be able

to provide greater per-tree benefits. In addition to questions of tree benefits, since tree cover is correlated with socioeconomic status in many U.S. cities (Schwarz et al. 2015), distribution of funding to a city’s wooded perimeter also raises questions of environmental justice. While concerns about the intra-city distribution of funding were not part of EABTP’s statewide policies, these topics likely would be relevant if similar programs were to be implemented on a municipal level.

Consistently high relative importance and structural value of white ash across all inventoried sites demonstrates the significance of these trees to the landscapes of program participants. While there were several sites where other species where more numerous, white ash ranked as the species with highest relative structural value on nine of sixteen properties inventoried. The consistent abundance and value of white ash across sites again suggests the interesting topic of identifying a threshold level of ash trees—or another threatened species—as a predictor of homeowner engagement in landscape tree preservation.

While white ash was present on all sites, green ash was represented by a single tree on one property. White and green ash are the two most widely distributed ash species in North

America; while they differ in site requirements, both were common statewide in Virginia forests before the arrival of EAB (Granger et al. 2017). Cultivars of both green and white ash were heavily planted as street trees from the 1940s to the 1990s (Poland and McCullough 2006), and

Urban Forest Assessment data from 2010-2011 shows that green ash were likely more abundant than white ash in Roanoke and Charlottesville. The current disparity between the species evident from on-site inventory data may reflect a difference in the species’ resistance to EAB infestation, which has been documented by multiple researchers (see Koch et al. 2015). The greater

susceptibility of green ash to EAB, then, could inform tree preservation initiatives—through preferential funding for green ash treatment, or other prioritization mechanisms.

While cluster analysis provided an interpretable typology of properties, the categories produced were not significant predictors in linear models of site species composition. Models of ash Relative importance and Relative structural value employed the predictors of cluster membership, Years since EAB infestation, and Historical ash abundance. None of these predictors were significant in either model, although Years since EAB infestation was marginally significant for the model of Relative structural value. This negative association is intuitive—as time has passed since a local initial EAB detection, the number and condition of ash trees has declined, leading to a decrease in value. With a greater sample size and a better-fitting model, this association could be used to model a decline in appraised value of ash trees for a single property over the course of an EAB infestation—information which might prove helpful to homeowners or insurers. Researchers have constructed city-wide models for a similar purpose, with the aim of informing decision makers (see Vannatta, Hauer, and Schuettpelz 2012, or Sadof et al. 2017). A parcel-level analysis of value lost through inaction or preserved by tree treatment could likewise assist homeowners in budgeting decisions.

While predictive models of site ash importance and structural value were not statistically significant here, this mode of analysis has been successfully employed by other researchers. An analysis of Boston, MA residential landscapes found associations between vegetation structure and property characteristics, include parcel area, year of construction, and architectural style

(Ossola et al. 2019). Schmitt-Harsh et al. (2013) described how tree species composition in

Bloomington, IN residential parcels varied by decade of property development—notably, white ash were commonly planted in that locality only in the 1950s and 1970s. A geospatial analysis of

urban vegetation diversity in Ballarat, Australia similarly found that a property’s physical characteristics outperformed socioeconomic characteristics in predicting tree cover and species richness (Kendal et al. 2012). Notably, in a 2017 study, researchers found greater success in using cluster analysis to classify street trees by land type or development era, compared to trees on residential properties (Nitoslawski et al. 2017). This type of parcel-level species composition modeling, while unlikely to be generalizable beyond a regional scale, could prove useful for municipal forest planning in the absence of detailed inventory data.

6.2.2 Homeowner Engagement in Landscape Tree Preservation

The relative homogeneity of respondent demographic characteristics demonstrates the relevance of broader outreach both for cost-share programming and for conservation research.

Statewide, about 77% of Virginia residents identify as White, while 19% identify as Black, and

3% as Asian (Wikipedia contributors 2019). Across racial groups, 9% identify as Hispanic (Pew

Research Center 2016). Respondents to the EABTP participants’ and general households survey were almost exclusively white and on average, in their late fifties. Differing rates of homeownership by race and ethnicity are unlikely to explain this disparity—at the national level, homeownership rates range only between 42% and 72% among ethnic groups (U. S. Census

Bureau 2017). For this reason, it is likely that there are sizeable populations of minority homeowners in Virginia who may be unaware of or not engaged with conservation funding such as the EABTP or conservation research, such as this project.

Additionally, younger homeowners were likely underrepresented among EABTP participants and among household survey respondents. Nationally, homeownership rates for those between 25 and 29 are just under 33%, and reach 57% for those between 35 and 39 (U. S.

100

Census Bureau 2017b). In designing the general households’ survey, a web+mail design was selected based on research indicating its effectiveness in increasing response rates among younger respondents (e.g., Sexton, Miller, and Dietsch 2011, de Bernardo and Curtis 2013).

Response data offered only minimal support for this hypothesis—while decreasing age was significantly associated with a web response, the effect was negligibly small. However, a choice of response mode is only one of many survey design aspects relevant to response rates. For researchers, other strategies for increasing minority response including contacting residents through community networks (Swanson and Ward 1995), or the use of snowball sampling (Perez et al. 2013). Low participation or response rates among younger and minority homeowners calls for a diversity of input and creative strategies to expand project outreach of cost-share programming and conservation research.

Models of homeowner Tree preservation intention and Cost-share participation demonstrated the consistent significance of attitudinal predictors, as opposed to externally observable characteristics. Tree preservation intention measured a respondent’s stated likelihood of investing personal funds in the long-term treatment of a threatened landscape tree, while Cost- share participation measured interest in a public subidy for tree treatment. For all top-ranking models of either dependent variable, the attitudinal predictor Importance of preserving ash/shade trees was significant or marginally significant. At first glance, this relationship is obvious—those who feel strongly about the importance of preserving landscape trees from pests are also the most likely to invest in tree treatment. However, this finding also demonstrates a useful principle for program planning—that strong supporters of urban forest health initiatives may reliably be enlisted from groups already involved in tree preservation. Tree stewardship groups and urban cooperative extension programs are two examples of organizations based on shared interest in

101

protection of urban natural resources (Day et al. 1997). Many members of these groups possess prior interest in tree preservation, familiarity with local urban forests, and social networks of their own. For these reasons, urban forest health initiatives could benefit from their involvement.

What is perhaps more surprising than the consistent importance of attitudinal predictors is the apparent lack of predictive power of most personal and property characteristics. Many of the characteristics tested here, such as lot size or the age of a house, are available through public databases, and for that reason have the potential to inform program outreach with few expenses upfront. Yet almost none of these variables, from household income to the number of trees on a property, showed any influence on the dependent variables of interest.

In constrast with the lack of findings here, the literature on cost-share enrollment among non-industrial private forest (NIPF) landowners provides many examples of the predictive power of personal or property characteristics. Studies have documented the significance of educational attainment (Watson et al. 2013), property size (Kline et al. 2000), and household income (Joshi and Arano 2009) as predictors of enrollment in cost-share programming. In a closer analogue to this research, Conway and Bang (2014) found that while household income, ethnicity, homeownership and canopy cover made no contribution to a model of support for a urban tree planting program, educational attainment and age of neighborhood development were predictive.

While in this research it is possible that relatively homogeneous samples may have limited statistical power, related findings indicate the possibility of constructing predictive models of homeowner decision-making using external characteristics.

Models of behavioral intention did identify one significant assocation apart from attitudinal predictors—a positive association between the Established EAB stratum and the willingness to pay for tree treatment among general households. In counties within this stratum,

102

EAB had been present for at least ten years. This finding indicates that personal awareness or experience of EAB damage may lead to greater willingness to invest in tree preservation. It also suggests the importance of communicating with residents living ahead of an impending invasion front, since they may be less likely to act without an immediate reason to do so. While this association may seem seem uncontroversial, there are confounding variables which weaken a causal claims. First, relative to the rest of the state, household income is high in many of the northern Virginia counties within the Established stratum. Secondly, relative to the entire state, ash trees were historically highly abundant in the same geographic region (see Figure 2).

Consequently, while it is interesting to note the association between local age of EAB infestation and willingness to pay for tree treatment, it is not possible to know, from these data, whether this finding would be generalizable across the state.

6.2.3 Practitioner Perceptions of Cost-Share Participation

Survey responses from both foresters and arborists represented most regions of Virginia, although no forester responses were recorded from southwestern Virginia. Response rate was low for foresters (18.1%) and even lower for arborists (9.1%), reducing the generalizability of study findings. While the final sample was also low for foresters (n=19), this group represented more than three-quarters of all county foresters who had participated in the 2018 program.

Interestingly, almost 75% of foresters reported having ISA certification, compared to a level of

10.5% for SAF certification. By itself, this statistic is an interesting finding—demonstrating the growing the relevance of this primarily arboricultural certification to the current work of VDOF county foresters. Additionally, this statistic suggests that EABTP site visits and approval of individual trees for treatment were not far removed from some foresters’ prior job duties.

103

The two top-ranking models for arborists’ interest in future EABTP participation included, separately, attitudinal and professional characteristics. Neither model was significant nor fit the data well. While of very little statistical value, the lone significant predictor was the binary variable EAB management recommendation cluster, in the model Professional

Characteristics. This result may indicate a greater likelihood of program interest among arborists with a preference for a ‘wait and see’ approach, when recommending a course of action to clients with ash trees, instead of preemptive tree removal. This result highlights the fact that some arborists may favor preemptive tree removal above all other options. Because dead ash trees are known to quickly become brittle, there are likely significant safety concerns tied to such a recommendation (Barnes et al. 2019). Not surprisingly, arborists expressing this preference were also unlikely to express strong interest in the EABTP. These results together demonstrate how arborists display a variety of attitudes and professional opinions regarding tree preservation and

EAB management. An awareness of these multiple points of view is valuable in enlisting arborist participation in future cost-share initiatives.

Finally, practitioners’ written responses regarding interest (or lack of interest) in the

EABTP help to further describe multiple viewpoints. About one-quarter of those responding indicated that EABTP was not relevant to their line of work, a figure that by itself reduces the generalizability of remaining responses. Of the remaining responses, the ratio of positive

(interested in EABTP participation) to negative (uninterested) was roughly 3:2. Considering that the work of foresters and arborists is essential to program implementation, this is a relatively narrow margin of support. Many of the negative responses regarded external conditions, such as lack of financial need or EAB awareness among clients, or simply a local lack of ash trees.

Others addressed respondents’ internal attitudes or opinions, such as lack of time for paperwork,

104

an opposition to chemical use, or the view that EABTP incentives were unsustainable. On the other hand, the two most common (non-neutral) responses overall were positive: the opinion that

EABTP was beneficial for ash preservation, and the opinion that the program was beneficial for clients. These two opinions align closely with EABTP’s program objectives of preserving populations of ash trees and ‘jumpstarting’ this work among Virginia residents (Chamberlin

2018a). These responses demonstrate that support for EABTP, while not universal, is strong among many practitioners.

6.3 Study Limitations

Strength of conclusions in parts of this research were limited by sample size. The number of participant property inventories (n=16) and the number of forester respondents (n=19) were below a commonly-accepted threshold of 30 necessary for the application of the central limit theorem (CLT) to non-normally distributed data (Howell 2013). For these datasets, statistical tests that rely on the CLT were avoided, except for linear models of ash importance and value on participant properties. These models, which were not significant, were presented as an exploratory exercise.

In addition to limitations of absolute sample size, strength of survey research conclusions were also limited by large margins of error in three out of four cases. Margin of error, or sampling error, is a calculated percent range within which a population mean is likely to fall, for a given confidence level (Vaske 2008). At the 95% confidence level, the calculated margin of error for the general households’ survey was ±4.84%, compared with larger values for the arborists’ survey (±7.79%), EABTP participants’ survey (±8.63%), and the foresters’ survey

(±20.4%).

105

Potential for nonresponse bias also affected the generalizability of some survey results.

For the general household survey, tests for nonresponse bias indicated that educational attainment of respondents was significantly higher than that of nonrespondents. The same tests also indicated marginally significant differences in household income level and the proportion of minorities. Taken together, nonresponse test results indicate that general household survey responses reflect a sample that is slightly more educated, wealthier, and more racially homogeneous than the sampling frame. For the arborists survey, tests for nonresponse bias also indicated that researchers or educators were overrepresented, and that landscape professionals and arborists in southeastern Virginia were underrepresented, compared with a reference group of MAC-ISA members statewide.

Another important limitation in analysis of tree inventory data was the timing of data collection. Since tree inventories were conducted after leaf drop, assessments of tree condition were less accurate than if they had been made during the growing season. Condition ratings were based on the relative absence of dead limbs, the relative absence of wounds or defects, and the relative presence of live buds on branch tips. However, these dormant season assessments necessarily made use of much less information than a visual assessment of tree canopy. A second, related limitation was a lack of accurate information on the treatment status of individual ash trees. Because much of the data collection was conducted without a homeowner present, there was often no clear way to determine which ash tree had been treated in 2018 or before.

Even when homeowners were present, not all were certain of the status of individual ash trees.

This gap in data collection prevented comparison of the condition of trees treated prior to 2018 to others treated in 2018, or untreated trees. It also precluded an analysis of factors that might influence the treatment of certain ash trees over others.

106

Survey instrument design limited the number of usable responses, and possibly the response rate itself. Because the general households’ survey was designed to capture responses from ash owners and non-ash owners separately, it contained two logical paths. On the paper version of the survey, the complexity of question structure may have contributed to a number of blank or unusable responses. Further, since the number of ash-owning households was so small

(n=10), these were excluded from analysis because much of the response data was not comparable to that of remaining households and did not constitute an adequate sample size on its own.

The topic of ash-owning households points to another significant limitation of homeowner surveys: the lack of an appropriate ‘control’ group of ash-owning households against which to compare EABTP participant responses. Comparison of participant responses to those of general households might prove interesting but is not valid for inferential testing because the circumstances faced by either group were completely different. Had a large number (>30) of ash- owning households been identified in the general household sample, a meaningful comparison could have been drawn between their interest in tree preservation and that of actual program participants. Potentially, this mode of analysis could identify predictors of ash preservation among a sample of specifically ash-owning households, and measure the potential influence of

EABTP cost-share assistance among a wider pool of eligible applicants. Instead, remaining households were asked about their intentions to preserve trees from a theoretical pest threat, and about their interest in a hypothetical cost-share program. While these responses were useful on their own, they are less directly applicable to the present EAB infestation.

Finally, the inclusion of ranked-choice items on survey instruments limited the types of possible analyses, particularly for practitioners’ surveys. These items had been included in

107

surveys for three reasons: (1) to reduce survey length by collapsing several Likert-scale items into a single ranking item, (2) to maintain respondent interest with varied item format, and (3) to gather information about relative preference that might not be gleaned from side-by-side Likert- scale items. While ranking items did serve the first purpose, and possibly the second, analysis of the resulting data yielded very little information, as judged by test results. Arborist and forester surveys contained ranking items regarding ash preservation, pest threat perceptions, and EAB management recommendations. From these items, each containing 5-7 ranking options, only one significant result was identified in further analysis. While it is possible ranking items could be improved with clearer wording and stronger separation between options, anecdotal evidence from this research suggests these items are a poor method of measuring preference.

6.4 Future Research

Many studies have been conducted on the effects of emamectin benzoate and other insecticides on EAB in controlled, experimental forests (e.g., Smitley, Doccola, and Cox 2010,

McCullough et al. 2011, McCullough et al. 2019). There are fewer examples of trials conducted in conducted in urban “field” conditions (e.g., Bick et al. 2018). While conducting trials with street and landscape trees adds many sources of variability, this experimental approach allows for testing of insecticidal performance in the stressful, urban conditions where most treated trees are located. Further testing of short- and long-term outcomes of insecticidal regimens for a variety of urban planting sites could benefit the hands-on work of tree preservation.

Further study of urban residential properties and households can help inform forest health initiatives in urban areas. The growing availability of urban FIA data (USFS 2014) may provide one avenue of pursuing this research. This USDA Forest Service program extends traditional

108

FIA plot-based, cyclical inventory into urban areas. In one example, researchers analyzed urban

FIA data and plot locations to map and measure levels of tree canopy inequity (Mills et al. 2016).

For locations where these data are available, neighborhood-level projections of tree species composition are possible. In contrast to this research, where households were selected based on relative tree cover, a survey sampling frame could be constructed at the neighborhood level using species-specific criteria from recent FIA data. This level of detailed urban inventory data has previously only been available for municipalities which have conducted plot-based tree inventories. Urban FIA opens possibilities for region- or nation-wide analyses of urban forest composition coupled with targeted household survey research.

Increasing availability of top-down urban forest assessments also creates possibilities for research and management of forest pests. As in the case of assessments conducted by the city

Milwaukee in 2009 (Sivyer 2010), hyperspectral imagery combined with LiDAR data allows for the identification of several tree genera with a accuracy of 80% or greater. In contrast with plot- based sampling, which helps estimate species composition of the surrounding area, remote sensing techniques have the potential to map out individual trees by genus on individual parcels.

Currently, the techniques involved are technically sophisticated and costly, and may be time- consuming conduct on a large scale. However, for locations where such assessments have already been conducted, fine-grained analyses of parcel-level tree inventory data coupled with household data are possible.

A related, but somewhat contrasting direction for research is the construction of models for tree species composition at the parcel level. In this research, linear models of site species composition as functions of property characteristics were poorly fitting and not significant, likely due to small sample size. Other research has demonstrated the feasibility of this approach—for

109

instance, modeling parcel carbon storage as a function of development era (Schmitt-Harsh et al.

2013) or parcel vegetation as a function of the owners’ lifestyle behavior (Grove et al. 2006).

These techniques, while likely to be highly region-specific, could potentially help urban foresters characterize urban vegetation across a city using few data inputs, and at very low cost.

Finally, further research is needed regarding attitudes of urban property owners towards tree preservation specifically, and urban ecosystems in general. This research has focused on identifying predictors of strong interest or disinterest in tree preservation and cost-share participation, with the intention of projecting the appeal of these practices broadly. A related direction for further study would focus on those with no strong inclination for or against tree preservation and identifying potential ‘nudge’ factors which might help increase the number of households supporting forest health objectives. Rather than financial incentives, ‘nudge’ factors include new policies or ways of presenting information that influence behavior (Kuhfuss et al.

2016). A comparison of the effectiveness of a financial incentive to normative influences of behavioral change could help inform methods of urban forest health outreach.

110

REFERENCES

Ajzen, Icek, and Martin Fishbein. 1980. Understanding Attitudes and Predicting Social

Behavior. Englewood Cliffs, N.J.: Prentice-Hall.

Alpert, Peter. 2006. “The Advantages and Disadvantages of Being Introduced.” Biological

Invasions 8 (7): 1523–34. https://doi.org/10.1007/s10530-005-5844-z.

Alvey, Alexis A. 2006. “Promoting and Preserving Biodiversity in the Urban Forest.” Urban

Forestry & Urban Greening 5 (4): 195–201. https://doi.org/10.1016/j.ufug.2006.09.003.

American Nurseryman. 2013. “Asian Longhorned Beetle Declared Eradicated in N.J., Canada,”

May 2013.

Andridge, Rebecca R, and Roderick J A Little. 2011. “Proxy Pattern-Mixture Analysis for

Survey Nonresponse.” Journal of Ofﬁcial Statistics, 27 (2): 153.

ArcGIS Pro (version 2.2). 2018. Redlands, CA: Environmental Systems Research Institute.

Arnold, Todd W. 2010. “Uninformative Parameters and Model Selection Using Akaike’s

Information Criterion.” The Journal of Wildlife Management 74 (6): 1175–78.

https://doi.org/10.1111/j.1937-2817.2010.tb01236.x.

Asaro, Chris. 2007. “Forest Health Review.” Charlottesville, VA: Virginia Department of

Forestry. http://www.dof.virginia.gov/infopubs/_fhr/FHR-2007-05_pub.pdf.

———. 2008. “Forest Health Review.” Charlottesville, VA: Virginia Department of Forestry.

http://www.dof.virginia.gov/infopubs/_fhr/FHR-2008-05_pub.pdf.

———. 2013. “Forest Health Report.” Charlottesville, VA: Virginia Department of Forestry.

http://www.dof.virginia.gov/infopubs/_fhr/FHR-2013-01_pub.pdf.

111

Asaro, Christopher, John T. Nowak, and Anthony Elledge. 2017. “Why Have Southern Pine

Beetle Outbreaks Declined in the Southeastern U.S. with the Expansion of Intensive Pine

Silviculture? A Brief Review of Hypotheses.” Forest Ecology and Management 391

(May): 338–48. https://doi.org/10.1016/j.foreco.2017.01.035.

Aukema, Juliann E., Brian Leung, Kent Kovacs, Corey Chivers, Kerry O. Britton, Jeffrey Englin,

Susan J. Frankel, et al. 2011. “Economic Impacts of Non-Native Forest Insects in the

Continental United States.” PLoS ONE 6 (9): e24587. https://doaj-org.ezproxy.lib.vt.edu.

Aukema, Juliann E., Deborah G. McCullough, Betsy Von Holle, Andrew M. Liebhold, Kerry

Britton, and Susan J. Frankel. 2010. “Historical Accumulation of Nonindigenous Forest

Pests in the Continental United States.” BioScience 60 (11): 886–97.

https://doi.org/10.1525/bio.2010.60.11.5.

Barden, Charles J., Stephen B. Jones, and Larry E. Biles. 1996. “Extension Forestry Education:

Reaching the People Who Make Decisions.” Journal of Forestry March: 31–35.

Barnes, Elizabeth, Cliff Sadof, and Lindsay Purcell. 2019. “Avoid Deadly Risk of Dying Ash

Trees with Timely Tree Removal.” https://www.purduelandscapereport.org/article/avoid-

deadly-risk-of-dying-ash-trees-with-timely-tree-removal/.

Bauer, Leah S., Jian J. Duan, Juli R. Gould, and Roy Van Driesche. 2015. “Progress in the

Classical Biological Control of Agrilus planipennis Fairmaire (Coleoptera: Buprestidae)

in North America.” The Canadian Entomologist 147 (03): 300–317.

https://doi.org/10.4039/tce.2015.18.

Bean, Meredith. 2019. “Unpublished Data.” Charlottesville, VA: Virginia Department of

Forestry.

112

Bernardo, Dana Huyser de, and Anna Curtis. 2013. “Using Online and Paper Surveys: The

Effectiveness of Mixed-Mode Methodology for Populations Over 50.” Research on

Aging 35 (2): 220–40. https://doi.org/10.1177/0164027512441611.

Bick, Emily N, Nora J Forbes, Christopher Haugen, Grant Jones, Shawn Bernick, and Fredric

Miller. 2018. “Seven-Year Evaluation of Insecticide Tools for Emerald Ash Borer in

Fraxinus pennsylvanica (Lamiales: Oleaceae) Trees.” Journal of Economic Entomology

111 (2): 732–40. https://doi.org/10.1093/jee/toy018.

Biemer, Paul, and Andy Peytchev. 2013. “Using Geocoded Census Data for Nonresponse Bias

Correction: An Assessment.” Journal of Survey Statistics and Methodology 1 (1): 24–44.

https://doi.org/10.1093/jssam/smt003.

Boyd, Roy G., Barbara J. Daniels, Robert Fallon, and William F. Hyde. 1988. “Measuring the

Effectiveness of Public Forestry Assistance Programs.” Forest Ecology and Management

23 (4): 297–309. https://doi.org/10.1016/0378-1127(88)90060-6.

Brasier, C. M. 1990. “China and the Origins of Dutch Elm Disease: An Appraisal.” Plant

Pathology 39 (1): 5–16. https://doi.org/10.1111/j.1365-3059.1990.tb02470.x.

Brooks, David J. 1985. “Public Policy and Long-Term Timber Supply in the South.” Forest

Science 31 (2): 342–57.

Caldas, Marcellus M., Jason S. Bergtold, Jeffrey M. Peterson, and Dietrich H. Earnhart. 2016.

“Land-Use Choices: The Case of Conservation Reserve Program (CRP) Re-Enrollment in

Kansas, USA.” Journal of Land Use Science 11 (5): 579–94.

https://doi.org/10.1080/1747423X.2016.1215563.

Campanella, Thomas J. 2003. Republic of Shade: New England and the American Elm. Yale

University Press.

113

Cappaert, David, Deborah G. McCullough, Therese M. Poland, and Nathan W. Siegert. 2005.

“Emerald Ash Borer in North America: A Research and Regulatory Challenge.”

American Entomologist 51 (3): 152–65. https://doi.org/10.1093/ae/51.3.152.

Carson, Richard T., Nicholas E. Flores, and Norman F. Meade. 2001. “Contingent Valuation:

Controversies and Evidence.” Environmental and Resource Economics 19 (2): 173–210.

https://doi.org/10.1023/A:1011128332243.

Chamberlin, Lori. 2018a. “Personal Correspondence.” Charlottesville, VA: Virginia Department

of Forestry.

———. 2018b. “Unpublished Data.” Charlottesville, VA: Virginia Department of Forestry.

Ciuzio, Elizabeth, William L. Hohman, Brian Martin, Mark D. Smith, Scott Stephens, Allan M.

Strong, and Tammy Vercauteren. 2013. “Opportunities and Challenges to Implementing

Bird Conservation on Private Lands.” Wildlife Society Bulletin 37 (2): 267–77.

https://doi.org/10.1002/wsb.266.

Clark, Stacy L., Scott E. Schlarbaum, Arnold M. Saxton, and Richard Baird. 2019. “Eight-Year

Blight (Cryphonectria parasitica) Resistance of Backcross-Generation American

Chestnuts (Castanea dentata) Planted in the Southeastern United States.” Forest Ecology

and Management 433 (February): 153–61. https://doi.org/10.1016/j.foreco.2018.10.060.

Conway, Tenley M., and Elizabeth Bang. 2014. “Willing Partners? Residential Support for

Municipal Urban Forestry Policies.” Urban Forestry & Urban Greening 13 (2): 234–43.

https://doi.org/10.1016/j.ufug.2014.02.003.

Costello, Anna B, and Jason W Osborne. 2005. “Best Practices in Exploratory Factor Analysis:

Four Recommendations for Getting the Most From Your Analysis.” Exploratory Factor

Analysis 10 (7): 9.

114

Cumbia, Dean. 2018. “Unpublished Data.” Charlottesville, VA: Virginia Department of Forestry.

Day, Eric R., Theresa A. Delinger, Douglas Pfeiffer, Mark Sutphin, and Chris Bergh. 2018. “Pest

Alert: Spotted Lanternfly, Lycorma deliculata.” Virginia Cooperative Extension.

https://vtechworks.lib.vt.edu/bitstream/handle/10919/87882/ENTO-291.pdf?sequence=1.

Day, Susan D., Sheri T. Dorn, Diane Relf, and Roger Harris. 1997. “Advanced Master Gardener

- Tree Steward: Training to Enhance Community Volunteerism".” HortTechnology 7 (4):

363–67. https://journals.ashs.org/horttech/view/journals/horttech/7/4/article-p363.pdf.

Dayer, Ashley A., Shorna Broussard Allred, and Richard C. Stedman. 2014. “Comparative

Analysis and Assessment of Forest Landowner Typologies.” Society & Natural

Resources 27 (11): 1200–1212. https://doi.org/10.1080/08941920.2014.933931.

Dillman, Don A., Jolene D. Smyth, and Leah Melani Christian. 2009. Internet, Mail, and Mixed-

Mode Surveys: The Tailored Design Method. Hoboken, NJ: John Wiley & Sons.

Dodds, Kevin J., and David A. Orwig. 2011. “An Invasive Urban Forest Pest Invades Natural

Environments - Asian Longhorned Beetle in Northeastern US Hardwood Forests.”

Canadian Journal of Forest Research 41: 1729–42.

Dolnicar, Sara. 2002. “A Review of Unquestioned Standards in Using Cluster Analysis for Data-

Driven Market Segmentation.” In Proceedings of the Australian and New Zealand

Marketing Academy Conference, 9. Deakin University: Melbourne.

Duan, Jian J., Leah S. Bauer, Roy G. van Driesche, and Juli R. Gould. 2018. “Progress and

Challenges of Protecting North American Ash Trees from the Emerald Ash Borer Using

Biological Control.” Forests 9 (3): 142. https://doi.org/10.3390/f9030142.

EAB Information Network. 2019. “Emerald Ash Borer Information Network.”

http://emeraldashborer.info/.

115

Egan, A., D. Gibson, and R. Whipkey. 2001. “Evaluating the Effectiveness of the Forest

Stewardship Program in West Virginia.” Journal of Forestry 99 (3): 31.

Esseks, J Dixon, and Robert J Moulton. 2000. “Evaluating the Forest Stewardship Program

Through a National Survey of Participating Forest Land Owners.” DeKalb, IL: Center for

Governmental Studies.

Farmer, James R., Zhao Ma, Michael Drescher, Eric G. Knackmuhs, and Stephanie L.

Dickinson. 2017. “Private Landowners, Voluntary Conservation Programs, and

Implementation of Conservation Friendly Land Management Practices.” Conservation

Letters 10 (1): 58–66. https://doi.org/10.1111/conl.12241.

Feng, Hongli, Catherine L. Kling, Lyubov A. Kurkalova, Silvia Secchi, and Philip W. Gassman.

2005. “The Conservation Reserve Program in the Presence of a Working Land

Alternative: Implications for Environmental Quality, Program Participation, and Income

Transfer.” American Journal of Agricultural Economics 87 (5): 1231–38.

https://doi.org/10.1111/j.1467-8276.2005.00812.x.

FIA. 2018. “Forest Statistics by State [Data File].” FIA DataMart FIADB_1.7.2.00. 2018.

https://apps.fs.usda.gov/fia/datamart/datamart_excel.html.

Goldman, Rebecca L., Barton H. Thompson, and Gretchen C. Daily. 2007. “Institutional

Incentives for Managing the Landscape: Inducing Cooperation for the Production of

Ecosystem Services.” Ecological Economics, Special Section - Ecosystem Services and

Agriculture, 64 (2): 333–43. https://doi.org/10.1016/j.ecolecon.2007.01.012.

Gormley, Isobel Claire, and Thomas Brendan Murphy. 2008. “Exploring Voting Blocs Within

the Irish Electorate: A Mixture Modeling Approach.” Journal of the American Statistical

Association 103 (483): 1014–27. https://doi.org/10.1198/016214507000001049.

116

Granger, Joshua J., John M. Zobel, and David S. Buckley. 2017. “Potential for Regenerating

Major and Minor Ash Species (Fraxinus spp.) Following EAB Infestation in the Eastern

United States.” Forest Ecology and Management 389 (April): 296–305.

https://doi.org/10.1016/j.foreco.2016.12.003.

Griffin, Gary J. 2000. “Blight Control and Restoration of the American Chestnut.” Journal of

Forestry; Bethesda 98 (2): 22.

https://search.proquest.com/docview/220789461/citation/114B6E95AE404B86PQ/1.

Grove, Jeremiah, Troy Arlington, J.P. O’Neil-Dunne, W.R. Burch, M.L Cadenasso, and S.T.A.

Pickett. 2006. “Characterization of Households and Its Implications for the Vegetation of

Urban Ecosystems.” Ecosystems 9 (4): 578–97.

https://www.fs.fed.us/ne/newtown_square/publications/other_publishers/ne_2006_grove0

02p.pdf.

Haack, R.A., Joseph F. Cavey, E. Richard Hobeke, and K Law. 1996. “Anoplophora

glabripennis: A New Tree-Infesting Exotic Cerambycid Invades New York.” Newsletter

of the Michigan Entomological Society, 1996.

Haack, Robert A. 2006. “Exotic Bark- and Wood-Boring Coleoptera in the United States: Recent

Establishments and Interceptions.” Canadian Journal of Forest Research 36 (2): 269.

Haack, Robert A., Franck Hérard, Jianghua Sun, and Jean J. Turgeon. 2010. “Managing Invasive

Populations of Asian Longhorned Beetle and Citrus Longhorned Beetle: A Worldwide

Perspective.” Annual Review of Entomology 55 (1): 521–46.

https://doi.org/10.1146/annurev-ento-112408-085427.

117

Hansen, Richard A., Amanda Clarke Henley, Emily S. Brouwer, Adaora N. Oraefo, and Mary T.

Roth. 2007. “Geographic Information System Mapping as a Tool to Assess Nonresponse

Bias in Survey Research.” Research in Social and Administrative Pharmacy 3 (3): 249–

64. https://doi.org/10.1016/j.sapharm.2006.10.001.

Hardie, Ian W., and Peter J. Parks. 1991. “Individual Choice and Regional Acreage Response to

Cost-Sharing in the South, 1971-1981.” Forest Science 37 (1): 175–90.

https://doi.org/10.1093/forestscience/37.1.175.

Herms, Daniel A., and Deborah G. McCullough. 2014. “Emerald Ash Borer Invasion of North

America: History, Biology, Ecology, Impacts, and Management.” Annual Review of

Entomology 59 (1): 13–30. https://doi.org/10.1146/annurev-ento-011613-162051.

Hippel, Paul von, David Hunter, and McKalie Drown. 2017. “Better Estimates from Binned

Income Data: Interpolated CDFs and Mean-Matching.” Sociological Science 4: 641–55.

https://doi.org/10.15195/v4.a26.

Howell, David C. 2013. Statistical Methods for Psychology. 8th ed. Belmont, CA: Wadsworth.

IBM SPSS for Windows, Version 25.0. 2017. Armonk, NY: IBM Corp.

Jacques, Julien, Quentin Grimonprez, and Christophe Biernacki. 2016. “Rankcluster: An R

Package for Clustering Multivariate Partial Rankings,” The R Journal 6(1): 10.

Jennings, David E., Jian J. Duan, Dick Bean, Juli R. Gould, Kimberly A. Rice, and Paula M.

Shrewsbury. 2016. “Monitoring the Establishment and Abundance of Introduced

Parasitoids of Emerald Ash Borer Larvae in Maryland, U.S.A.” Biological Control 101

(October): 138–44. https://doi.org/10.1016/j.biocontrol.2016.07.006.

118

Jennings, David E., Philip B. Taylor, and Jian J. Duan. 2014. “The Mating and Oviposition

Behavior of the Invasive Emerald Ash Borer (Agrilus planipennis), with Reference to the

Influence of Host Tree Condition.” Journal of Pest Science 87 (1): 71–78.

https://doi.org/10.1007/s10340-013-0539-1.

Joshi, Sudiksha, and Kathryn G. Arano. 2009. “Determinants of Private Forest Management

Decisions: A Study on West Virginia NIPF Landowners.” Forest Policy and Economics

11 (2): 118–25. https://doi.org/10.1016/j.forpol.2008.10.005.

Kappler, Rachel H., Kathleen S. Knight, Jennifer Koch, and Karen V. Root. 2018. “Neighboring

Tree Effects and Soil Nutrient Associations with Surviving Green Ash (Fraxinus

pennsylvanica) in an Emerald Ash Borer (Agrilus planipennis) Infested Floodplain

Forest.” Forests 9 (4): 183. https://doi.org/10.3390/f9040183.

Kendal, Dave, Nicholas S. G. Williams, and Kathryn J. H. Williams. 2012. “Drivers of Diversity

and Tree Cover in Gardens, Parks and Streetscapes in an Australian City.” Urban

Forestry & Urban Greening 11 (3): 257–65. https://doi.org/10.1016/j.ufug.2012.03.005.

Kilgore, Michael A., John L. Greene, Michael G. Jacobson, Thomas J. Straka, and Steven E.

Daniels. 2007a. “The Influence of Financial Incentive Programs in Promoting Sustainable

Forestry on the Nation’s Family Forests.” Journal of Forestry 105 (4): 184–91.

https://doi.org/10.1093/jof/105.4.184.

Kline, Jeffrey D., Ralph J. Alig, and Rebecca L. Johnson. 2000. “Fostering the Production of

Nontimber Services Among Forest Owners with Heterogeneous Objectives.” Forest

Science 46 (2): 302–11. https://doi.org/10.1093/forestscience/46.2.302.

Kline, Jeffrey D, and Brett J Butler. 2002. “Tree Planting in the South: What Does the Future

Hold?” Southern Journal of Applied Forestry 26 (2): 99–107.

119

Klooster, Wendy, Kamal Gandhi, Lawrence Long, Kayla Perry, Kevin Rice, and Daniel Herms.

2018. “Ecological Impacts of Emerald Ash Borer in Forests at the Epicenter of the

Invasion in North America.” Forests 9 (5): 250. https://doi.org/10.3390/f9050250.

Knight, Kathleen S., John P. Brown, and Robert P. Long. 2013. “Factors Affecting the Survival

of Ash Trees Infested by Emerald Ash Borer.” Biological Invasions 15 (2): 371–83.

https://doi.org/10.1007/s10530-012-0292-z.

Knight, Kathleen S, Daniel Herms, Reid Plumb, and Eileen Sawyer. 2011. “Dynamics of

Surviving Ash (Fraxinus spp.) Populations in Areas Long Infested by Emerald Ash Borer

(Agrilus planipennis).” In 4th International Workshop on Genetics of Host-Parasite

Interactions in Forestry, 10.

Koch, J. L., D. W. Carey, M. E. Mason, T. M. Poland, and K. S. Knight. 2015. “Intraspecific

Variation in Fraxinus pennsylvanica Responses to Emerald Ash Borer (Agrilus

planipennis).” New Forests 46 (5–6): 995–1011. https://doi.org/10.1007/s11056-015-

9494-4.

Kokotovich, Adam E., and Adam R. Zeilinger. 2011. “Exploring Social and Institutional

Variation across Oak Wilt Risk Management Programs in Minnesota, USA.” Urban

Forestry & Urban Greening 10 (1): 39–45. https://doi.org/10.1016/j.ufug.2010.09.004.

Komen, James, and Donald R Hodel. 2015. “An Analysis of the Field Precision of the CTLA

Trunk Formula Method,” Arboriculture & Urban Forestry 41 (5) 279-285.

Kondo, Michelle, Jaime Fluehr, Thomas McKeon, and Charles Branas. 2018. “Urban Green

Space and Its Impact on Human Health.” International Journal of Environmental

Research and Public Health 15 (3): 445. https://doi.org/10.3390/ijerph15030445.

120

Kovacs, Kent F., Robert G. Haight, Deborah G. McCullough, Rodrigo J. Mercader, Nathan W.

Siegert, and Andrew M. Liebhold. 2010. “Cost of Potential Emerald Ash Borer Damage

in U.S. Communities, 2009–2019.” Ecological Economics 69 (3): 569–78.

https://doi.org/10.1016/j.ecolecon.2009.09.004.

Kovacs, Kent F., Robert G. Haight, Rodrigo J. Mercader, and Deborah G. McCullough. 2014. “A

Bioeconomic Analysis of an Emerald Ash Borer Invasion of an Urban Forest with

Multiple Jurisdictions.” Resource and Energy Economics 36 (1): 270–89.

https://doi.org/10.1016/j.reseneeco.2013.04.008.

Kuhfuss, Laure, Raphaële Préget, Sophie Thoyer, Nick Hanley, Philippe Le Coent, and Mathieu

Désolé. 2016. “Nudges, Social Norms, and Permanence in Agri-Environmental

Schemes.” Land Economics 92 (4): 641–55. https://doi.org/10.3368/le.92.4.641.

Lal, Pankaj, Bernabas Wolde, Janaki Alavalapati, Pralhad Burli, and John Munsell. 2016.

“Forestland Owners’ Willingness to Plant Pine on Non-Forested Land for Woody

Bioenergy in Virginia.” Forest Policy and Economics 73 (December): 52–57.

https://doi.org/10.1016/j.forpol.2016.09.003.

Lee, Jun-Hak, Yekang Ko, and E. Gregory McPherson. 2016. “The Feasibility of Remotely

Sensed Data to Estimate Urban Tree Dimensions and Biomass.” Urban Forestry &

Urban Greening 16 (January): 208–20. https://doi.org/10.1016/j.ufug.2016.02.010.

Lesser, Virginia M, and Lydia D Newton. 2016. “Mixed-Mode Surveys Compared With Single-

Mode Surveys: Trends in Responses and Methods to Improve Completion.” Journal of

Rural Social Sciences 31 (6): 7–34.

Li, Yanshu, and Daowei Zhang. 2007. “A Spatial Panel Data Analysis of Tree Planting in the US

South.” Southern Journal of Applied Forestry, 31 (4): 192-198.

121

Liebhold, A., V. Mastro, and P. w. Schaefer. 1989. “Learning from the Legacy of Leopold

Trouvelot.” Bulletin of the Entomological Society of America 35 (2): 20–22.

https://doi.org/10.1093/besa/35.2.20.

Liebhold, Andrew M., William L. MacDonald, Dale Bergdahl, and Victor C. Mastro. 1995

“Invasion by Exotic Forest Pests: A Threat to Forest Ecosystems.” Forest Science

41(suppl_1).

Liebhold, Andrew M. 2012. “Forest Pest Management in a Changing World.” International

Journal of Pest Management 58 (3): 289–95.

https://doi.org/10.1080/09670874.2012.678405.

Liu, Houping. 2013. “Emerald Ash Borer Management Plan for Pennsylvania Communities.”

Harrisburg, PA: Pennsylvania Department of Conservation and Natural Resources.

http://www.docs.dcnr.pa.gov/cs/groups/public/documents/document/dcnr_20027832.pdf.

———. 2018. “Under Siege: Ash Management in the Wake of the Emerald Ash Borer.” Journal

of Integrated Pest Management 9 (1). https://doi.org/10.1093/jipm/pmx029.

Liu, Houping, Leah S. Bauer, Deborah L. Miller, Tonghai Zhao, Ruitong Gao, Liwen Song,

Qingshu Luan, Ruozhong Jin, and Changqi Gao. 2007. “Seasonal Abundance of Agrilus

planipennis (Coleoptera: Buprestidae) and Its Natural Enemies Oobius agrili

(Hymenoptera: Encyrtidae) and Tetrastichus planipennisi (Hymenoptera: Eulophidae) in

China.” Biological Control 42 (1): 61–71.

https://doi.org/10.1016/j.biocontrol.2007.03.011.

122

Loo, Judy A. 2009. “Ecological Impacts of Non-Indigenous Invasive Fungi as Forest

Pathogens.” In Ecological Impacts of Non-Native Invertebrates and Fungi on Terrestrial

Ecosystems, edited by David W. Langor and Jon Sweeney, 81–96. Dordrecht: Springer

Netherlands. https://doi.org/10.1007/978-1-4020-9680-8_6.

Lovett, Gary M., Marissa Weiss, Andrew M. Liebhold, Thomas P. Holmes, Brian Leung, Kathy

Fallon Lambert, David A. Orwig, et al. 2016. “Nonnative Forest Insects and Pathogens in

the United States: Impacts and Policy Options.” Ecological Applications 26 (5): 1437–55.

https://doi.org/10.1890/15-1176.

Luzar, E. Jane, and Assane Diagne. 1999. “Articles: Participation in the next Generation of

Agriculture Conservation Programs: The Role of Environmental Attitudes.” Journal of

Socio-Economics 28 (January): 335–49. https://doi.org/10.1016/S1053-5357(99)00021-9.

MacFarlane, David W., and Shawna Patterson Meyer. 2005. “Characteristics and Distribution of

Potential Ash Tree Hosts for Emerald Ash Borer.” Forest Ecology and Management 213

(1): 15–24. https://doi.org/10.1016/j.foreco.2005.03.013.

McCullough, Deborah G., and Rodrigo J. Mercader. 2012. “Evaluation of Potential Strategies to

SLow Ash Mortality (SLAM) Caused by Emerald Ash Borer (Agrilus Planipennis ):

SLAM in an Urban Forest.” International Journal of Pest Management 58 (1): 9–23.

https://doi.org/10.1080/09670874.2011.637138.

McCullough, Deborah G., Therese M. Poland, Andrea C. Anulewicz, Phillip Lewis, and David

Cappaert. 2011. “Evaluation of Agrilus planipennis (Coleoptera: Buprestidae) Control

Provided by Emamectin Benzoate and Two Neonicotinoid Insecticides, One and Two

Seasons After Treatment.” Journal of Economic Entomology 104 (5): 1599–1612.

https://doi.org/10.1603/EC11101.

123

McCullough, Deborah G, Therese M Poland, Andrew R Tluczek, Andrea Anulewicz, James

Wieferich, and Nathan W Siegert. 2019. “Emerald Ash Borer (Coleoptera: Buprestidae)

Densities Over a 6-Yr Period on Untreated Trees and Trees Treated With Systemic

Insecticides at 1-, 2-, and 3-Yr Intervals in a Central Michigan Forest.” Journal of

Economic Entomology 112 (1): 201–12. https://doi.org/10.1093/jee/toy282.

McFadden, Max W., and Michael E. McManus. 1991. “An Insect out of Control? The Potential

for Spread and Establishment of the Gypsy Moth in New Forest Areas in the United

States.” In Forest Insect Guilds: Patterns of Interaction with Host Trees, 153:172–86.

Radno, PA: USDA, Northeastern Forest Experiment Station.

Mcmanus, Michael, and György Csóka. 2007. “History and Impact of Gypsy Moth in North

America and Comparison to Recent Outbreaks in Europe,” 18.

MD DNR-FS. 2017. “Emerald Ash Borer Treatment Program Description and Application for

Cost Share and Payment.” Maryland Department of Natural Resources-Forest Service.

http://dnr.maryland.gov/forests/Documents/Emerald-Ash-Borer-Treatment-Program-

Description-and-Application.pdf.

Medway, R. L., and J. Fulton. 2012. “When More Gets You Less: A Meta-Analysis of the Effect

of Concurrent Web Options on Mail Survey Response Rates.” Public Opinion Quarterly

76 (4): 733–46. https://doi.org/10.1093/poq/nfs047.

Mills, John R., Patrick Cunningham, and Geoffrey H. Donovan. 2016. “Urban Forests and Social

Inequality in the Pacific Northwest.” Urban Forestry & Urban Greening 16: 188–96.

https://doi.org/10.1016/j.ufug.2016.02.011.

124

Moll, Gary. 1993. “A New Way to See Our City Forests.” American Forests 99 (9–10): 29-.

http%3A%2F%2Flink.galegroup.com%2Fapps%2Fdoc%2FA14536163%2FAONE%3Fu

%3Dviva_vpi%26sid%3DAONE%26xid%3D454cd460.

Müllensiefen, Daniel, Christian Hennig, and Hedie Howells. 2018. “Using Clustering of

Rankings to Explain Brand Preferences with Personality and Socio-Demographic

Variables.” Journal of Applied Statistics 45 (6): 1009–29.

https://doi.org/10.1080/02664763.2017.1339025.

NCFS. 2018a. “Ash Protection Program Application.” North Carolina Forest Service.

http://www.ncforestservice.gov/forest_health/LSR2018/EABApplicationpart1.pdf.

———. 2018b. “Ash Protection Program Guidelines.” North Carolina Forest Service.

http://www.ncforestservice.gov/forest_health/LSR2018/LSR_Grant_Program%20Guideli

nes.pdf.

Nitoslawski, Sophie A., James W. Steenberg, Peter N. Duinker, and Peter G. Bush. 2017.

“Assessing the Influence of Location Attributes on Urban Forest Species Composition in

Suburban Neighbourhoods.” Urban Forestry & Urban Greening 27 (October): 187–95.

https://doi.org/10.1016/j.ufug.2017.08.002.

Nowak, David J., Eric J. Greenfield, Robert E. Hoehn, and Elizabeth Lapoint. 2013. “Carbon

Storage and Sequestration by Trees in Urban and Community Areas of the United

States.” Environmental Pollution 178 (July): 229–36.

https://doi.org/10.1016/j.envpol.2013.03.019.

Nowak, David J., Judith E. Pasek, Ronaldo A. Sequeira, Daniel E. Crane, and Victor C. Mastro.

2001. “Potential Effect of Anoplophora glabripennis (Coleoptera: Cerambycidae) on

125

Urban Trees in the United States.” Journal of Economic Entomology 94 (1): 116–22.

https://doi.org/10.1603/0022-0493-94.1.116.

Nowak, John, Christopher Asaro, Kier Klepzig, and Ronald Billings. 2008. “The Southern Pine

Beetle Prevention Initiative: Working for Healthier Forests.” Journal of Forestry;

Bethesda 106 (5): 261–67. https://search-proquest-

com.ezproxy.lib.vt.edu/docview/220812955/abstract/90FBFA1902154B17PQ/1.

NRCS. 2019. “Native, Invasive, and Other Plant-Related Definitions.” Natural Resources

Conservation Service. 2019.

https://www.nrcs.usda.gov/wps/portal/nrcs/detail/ct/technical/ecoscience/invasive/?cid=n

rcs142p2_011124.

Ossola, Alessandro, Dexter Locke, Brenda Lin, and Emily Minor. 2019. “Greening in Style:

Urban Form, Architecture and the Structure of Front and Backyard Vegetation.”

Landscape and Urban Planning 185 (May): 141–57.

https://doi.org/10.1016/j.landurbplan.2019.02.014.

Paap, Trudy, Treena I. Burgess, and Michael J. Wingfield. 2017. “Urban Trees: Bridge-Heads

for Forest Pest Invasions and Sentinels for Early Detection.” Biological Invasions 19

(12): 3515–26. https://doi.org/10.1007/s10530-017-1595-x.

Paillet, Frederick L. 2002. “Chestnut: History and Ecology of a Transformed Species.” Journal

of Biogeography 29 (10–11): 1517–30. https://doi.org/10.1046/j.1365-

2699.2002.00767.x.

Pautasso, Marco, Markus Schlegel, and Ottmar Holdenrieder. 2015. “Forest Health in a

Changing World.” Microbial Ecology 69 (4): 826–42. https://doi.org/10.1007/s00248-

014-0545-8.

126

Perez, Daniel F., Jason X. Nie, Chris I. Ardern, Natasha Radhu, and Paul Ritvo. 2013. “Impact of

Participant Incentives and Direct and Snowball Sampling on Survey Response Rate in an

Ethnically Diverse Community: Results from a Pilot Study of Physical Activity and the

Built Environment.” Journal of Immigrant and Minority Health 15 (1): 207–14.

https://doi.org/10.1007/s10903-011-9525-y.

Pew Research Center. 2016. “Latinos in the 2016 Election: Virginia.” 2016.

https://www.pewhispanic.org/fact-sheet/latinos-in-the-2016-election-virginia/.

Pimentel, David, Rodolfo Zuniga, and Doug Morrison. 2005. “Update on the Environmental and

Economic Costs Associated with Alien-Invasive Species in the United States.”

Ecological Economics, Integrating Ecology and Economics in Control Bioinvasions, 52

(3): 273–88. https://doi.org/10.1016/j.ecolecon.2004.10.002.

Poland, Therese M., and Deborah G. McCullough. 2006. “Emerald Ash Borer: Invasion of the

Urban Forest and the Threat to North America’s Ash Resource.” Journal of Forestry;

Bethesda 104 (3): 118–24. https://search-proquest-

com.ezproxy.lib.vt.edu/docview/220779905/abstract/8DA0E9356E204A31PQ/1.

Qualtrics. 2018. Qualtrics. Provo, UT.

R Core Team. 2018. R: A Language and Environment for Statistical Computing (version 3.5.2).

Vienna: R Foundation for Statistical Computing. http://www.R-project.org/.

Richardson, David M, and Petr Pyšek. 2004. “What Is an Invasive Species?” Crop Protection

Compendium, CAB International, Wallingford.

http:/www. cabicompendium. org/cpc

127

Rigling, Daniel, and Simone Prospero. 2018. “Cryphonectria parasitica, the Causal Agent of

Chestnut Blight: Invasion History, Population Biology and Disease Control.” Molecular

Plant Pathology 19 (1): 7–20. https://doi.org/10.1111/mpp.12542.

RStudio Team. 2019. Rstudio: Integrated Development for R. Boston, MA: RStudio, Inc.

http://www.rstudio.com.

Ryan, Richard M, and Edward L Deci. 2000. “Self-Determination Theory and the Facilitation of

Intrinsic Motivation, Social Development, and Well-Being.” American Psychologist, 11.

Sadof, Clifford S., Gabriel P. Hughes, Adam R. Witte, Donnie J. Peterson, and Matthew D.

Ginzel. 2017. “Tools for Staging and Managing Emerald Ash Borer in the Urban Forest.”

Arboriculture & Urban Forestry 43 (1): 15–26.

Sadof, Clifford S., Lindsey Purcell, Forrest J. Bishop, Carlos Quesada, and Zhi-Wei Zhan. 2011.

“Evaluating Restoration Capacity and Costs of Managing the Emerald Ash Borer with a

Web-Based Cost Calculator in Urban Forests.” Arboriculture & Urban Forestry 37 (2):

74–83.

Schlarbaum, Scott E, F Hebard, P.C. Spaine, and J.C. Kamalay. 1997. “Three American

Tragedies: Chestnut Blight, Butternut Canker, and Dutch Elm Disease.” In Exotic Pests

of Eastern Forests, 10. Nashville: Tennesse Exotic Pest Plant Council.

Schmiedel, Doreen, Eckehard-G. Wilhelm, Mechthild Roth, Cornelia Scheibner, Stefan Nehring,

and Susanne Winter. 2016. “Evaluation System for Management Measures of Invasive

Alien Species.” Biodiversity and Conservation 25 (2): 357–74.

https://doi.org/10.1007/s10531-016-1054-5.

128

Schmitt-Harsh, Mikaela, Sarah K. Mincey, Matt Patterson, Burnell C. Fischer, and Tom P.

Evans. 2013. “Private Residential Urban Forest Structure and Carbon Storage in a

Moderate-Sized Urban Area in the Midwest, United States.” Urban Forestry & Urban

Greening 12 (4): 454–63. https://doi.org/10.1016/j.ufug.2013.07.007.

Schwarz, Kirsten, Michail Fragkias, Christopher G. Boone, Weiqi Zhou, Melissa McHale, J.

Morgan Grove, Jarlath O’Neil-Dunne, et al. 2015. “Trees Grow on Money: Urban Tree

Canopy Cover and Environmental Justice.” PLOS ONE 10 (4): e0122051.

https://doi.org/10.1371/journal.pone.0122051.

Sexton, Natalie R., Holly M. Miller, and Alia M. Dietsch. 2011. “Appropriate Uses and

Considerations for Online Surveying in Human Dimensions Research.” Human

Dimensions of Wildlife 16 (3): 154–63. https://doi.org/10.1080/10871209.2011.572142.

Sharov, A. A., D. Leonard, A. M. Liebhold, E. A. Roberts, and W. Dickerson. 2002. “‘Slow The

Spread’: A National Program to Contain the Gypsy Moth.” Journal of Forestry 100 (5):

30–36. https://doi.org/10.1093/jof/100.5.30.

Siegert, Nathan W., Deborah G. McCullough, Andrew M. Liebhold, and Frank W. Telewski.

2014. “Dendrochronological Reconstruction of the Epicentre and Early Spread of

Emerald Ash Borer in North America.” Diversity and Distributions 20 (7): 847–58.

https://doi.org/10.1111/ddi.12212.

Sivyer, David. 2010. “Mapping the Future for EAB Readiness and Response Planning in

Milwaukee: An Update.” City Trees January: 16.

Smith, Michael T., Jay Bancroft, Guohong Li, Ruitong Gao, and Stephen Teale. 2001. “Dispersal

of Anoplophora glabripennis (Cerambycidae).” Environmental Entomology 30 (6):

1036–40. https://doi.org/10.1603/0046-225X-30.6.1036.

129

Smitley, David, Terrance Davis, and Eric Rebek. 2008. “Progression of Ash Canopy Thinning

and Dieback Outward from the Initial Infestation of Emerald Ash Borer (Coleoptera:

Buprestidae) in Southeastern Michigan.” Journal of Economic Entomology 101 (5):

1643–50. https://doi.org/10.1603/0022-0493(2008)101[1643:POACTA]2.0.CO;2.

Smitley, David R, Joseph J Doccola, and David L Cox. 2010. “Multiple-Year Protection of Ash

Trees from Emerald Ash Borer with a Single Trunk Injection of Emamectin Benzoate,

and Single-Year Protection with an Imidacloprid Basal Drench.” Arboriculture & Urban

Forestry 36 (5): 206–11.

Sorice, Michael G. 2012. “Retooling the Traditional Approach to Studying the Belief–Attitude

Relationship: Explaining Landowner Buy-In to Incentive Programs.” Society & Natural

Resources 25 (5): 499–512. https://doi.org/10.1080/08941920.2011.603143.

Sorice, Michael G., and J. Richard Conner. 2010. “Predicting Private Landowner Intentions to

Enroll in an Incentive Program to Protect Endangered Species.” Human Dimensions of

Wildlife 15 (2): 77–89. https://doi.org/10.1080/10871200903551985.

Soto, José R., Francisco J. Escobedo, Hayk Khachatryan, and Damian C. Adams. 2018.

“Consumer Demand for Urban Forest Ecosystem Services and Disservices: Examining

Trade-Offs Using Choice Experiments and Best-Worst Scaling.” Ecosystem Services 29

(February): 31–39. https://doi.org/10.1016/j.ecoser.2017.11.009.

Spector, Tova, and Francis E. Putz. 2006. “Crown Retreat of Open-Grown Southern Live Oaks

(Quercus virginiana) Due to Canopy Encroachment in Florida, USA.” Forest Ecology

and Management 228 (1): 168–76. https://doi.org/10.1016/j.foreco.2006.03.001.

130

Sun, Changyou. 2007. “Variation of Federal Cost-Share Programs in the United States and the

Inducement Effects on Tree Planting.” Journal of Forest Economics 12 (4): 279–96.

https://doi.org/10.1016/j.jfe.2006.09.002.

Swanson, G. M., and A. J. Ward. 1995. “Recruiting Minorities Into Clinical Trials Toward a

Participant-Friendly System.” JNCI Journal of the National Cancer Institute 87 (23):

1747–59. https://doi.org/10.1093/jnci/87.23.1747.

Symonds, Matthew R. E., and Adnan Moussalli. 2011. “A Brief Guide to Model Selection,

Multimodel Inference and Model Averaging in Behavioural Ecology Using Akaike’s

Information Criterion.” Behavioral Ecology and Sociobiology 65 (1): 13–21.

https://doi.org/10.1007/s00265-010-1037-6.

Theoharides, Kathleen A., and Jeffrey S. Dukes. 2007. “Plant Invasion across Space and Time:

Factors Affecting Nonindigenous Species Success during Four Stages of Invasion.” The

New Phytologist 176 (2): 256–73.

http://www.jstor.org.ezproxy.lib.vt.edu/stable/4627161.

Train, Kenneth E. 1998. “Recreation Demand Models with Taste Differences over People.” Land

Economics 74 (2): 230–39. https://doi.org/10.2307/3147053.

U.S. Census Bureau. 2010. “2010 Geographic Terms and Concepts - Block Groups.”

https://www.census.gov/geo/reference/gtc/gtc_bg.html.

———. 2012. “Qualifying Urban Areas for the 2010 Census.” Federal Register 77(59): 18652.

———. 2016. “New Census Data Show Differences Between Urban and Rural Populations.”

The United States Census Bureau. 2016. https://www.census.gov/newsroom/press-

releases/2016/cb16-210.html.

131

———. 2017a. “2016 American Community Survey 5-Year Estimates.”

https://factfinder.census.gov/faces/nav/jsf/pages/download_center.xhtml.

———. 2017b. “Housing Vacancies and Homeownership (CPS/HVS).”

https://www.census.gov/housing/hvs/data/ann17ind.html.

———. 2018. “TIGER Products.” https://www.census.gov/geo/maps-data/data/tiger.html.

USDA-APHIS. 2017. “Emerald Ash Borer Biological Control Release and Recovery

Guidelines.” Riverdale, MD: USDA-Animal and Plant Health Inspection Service-

Agricultural Research Service-Forest Service.

https://www.aphis.usda.gov/plant_health/plant_pest_info/emerald_ash_b/downloads/EA

B-FieldRelease-Guidelines.pdf.

USFS. 2014. “Urban FIA Brochure.” USDA Forest Service.

https://www.nrs.fs.fed.us/fia/urban/docs/Urban%20FIA%20Brochure%202014-opt.pdf.

———. 2018a. “FIA EVALIDator Version 1.8.0.00.” USDA Forest Service.

https://apps.fs.usda.gov/Evalidator/evalidator.jsp.

———. 2018b. “I-Tree Eco.” USDA Forest Service. https://www.itreetools.org/eco/.

———. 2019a. “Urban FIA Program.” USDA Forest Service.

https://www.nrs.fs.fed.us/fia/urban/.

———. 2019b. “Urban Natural Resources Stewardship.” USDA Forest Service - Northern

Research Station. https://www.nrs.fs.fed.us/urban/.

Van Driesche, Roy G., and Richard C. Reardon. 2015. “Biology and Control of Emerald Ash

Borer.” Morgantown, WV: USDA-Forest Service-Forest Health Technology Enterprise

Team. https://permanent.access.gpo.gov/gpo59706/FHTET-2014-

09_Biology_Control_EAB.pdf.

132

Vannatta, A. R., R. H. Hauer, and N. M. Schuettpelz. 2012. “Economic Analysis of Emerald Ash

Borer (Coleoptera: Buprestidae) Management Options.” Journal of Economic

Entomology 105 (1): 196–206. https://doi.org/10.1603/EC11130.

Vaske, Jerry J. 2008. Survey Research and Analysis: Application in Parks, Recreation and

Human Dimensions. State College, PA: Venture Publishing, Inc.

VDOF. 2018a. “Emerald Ash Borer Treatment Program Application.” Form 6.5. Charlottesville,

VA: Virginia Department of Forestry. http://www.dof.virginia.gov/forms/f06.05.dotx.

———. 2018b. “Forestry Topic 32 - Emerald Ash Borer Cost-Share Program.” FT0032.

Charlottesville, VA: Virginia Department of Forestry.

———. 2018c. “Policy and Procedure - Emerald Ash Borer Treatment Program.” PP_06-002.

Charlottesville, VA: Virginia Department of Forestry.

———. 2018d. “Policy and Procedure 6-2 Emerald Ash Borer Cost-Share Program.”

Charlottesville, VA: Virginia Department of Forestry.

VGIN. 2018a. “VGIN Virginia GIS Clearinghouse.” Virginia Geographic Information Network -

GIS Clearinghouse. 2018. https://vgin.maps.arcgis.com/home/index.html.

———. 2018b. “Virginia Land Cover Dataset.”

https://vgin.maps.arcgis.com/home/item.html?id=d3d51bb5431a4d26a313f586c7c2c848.

Virginia Tech. 2018. “VT Urban Forestry Municipal Urban Forest Assessments.”

https://www.urbanforestry.frec.vt.edu/eco.shtml.

Watson, Adam C., Jay Sullivan, Gregory S. Amacher, and Christopher Asaro. 2013. “Cost

Sharing for Pre-Commercial Thinning in Southern Pine Plantations: Willingness to

Participate in Virginia’s Pine Bark Beetle Prevention Program.” Forest Policy and

Economics 34 (September): 65–72. https://doi.org/10.1016/j.forpol.2013.05.004.

133

Wikipedia contributors. 2019. “Demographics of Virginia.” In Wikipedia, The Free

Encyclopedia.

https://en.wikipedia.org/w/index.php?title=Demographics_of_Virginia&oldid=88460590

Williamson, Mark. 1996. Biological Invasions. London: Chapman and Hall.

Wright, G Tyler. 2011. “An Assessment of the Potential Impacts of Emerald Ash Borer (Agrilus

planipennis Fairmaire) on Virginia’s Municipal Street Trees.” Virginia Tech.

Yong, An Gie, and Sean Pearce. 2013. “A Beginner’s Guide to Factor Analysis: Focusing on

Exploratory Factor Analysis.” Tutorials in Quantitative Methods for Psychology 9 (2):

79–94. https://doi.org/10.20982/tqmp.09.2.p079.

Zhang, Daowei. 1996. “Differences in Silvicultural Investment Under Various Types of Forest

Tenure in British Columbia.” Forest Science 42 (4): 442–49.

134

APPENDIX A Supplementary Tables for Chapter 3: Urban Forest Composition of Participant Properties

List of Tables

Table 20. List of Urban Participant Properties (UPP) paired with nearest Urban Forest Assessment (UFA) municipality ...... 136

Table 21. Summary of species composition variables for on-site inventories of Urban Participant Properties ...... 136

Table 22. Comparisons of inventoried and non-inventoried Urban Participant Properties ...... 137

Table 23. Summary of ash species composition, sampling methods, and relative error for Urban Forest Assessments, by municipality ...... 137

Table 24. Tree benefits by species across 16 Urban Participant Properties ...... 138

Table 25. Summary of site characteristics and species composition of inventoried Urban Participant Properties ...... 139

135

Table 20. List of Urban Participant Properties (UPP) paired with nearest Urban Forest Assessment (UFA) municipality. Sites within 15 mi (24 km) of UFA centroid were used in analysis. Nearest UFA Distance from UPP to UPP site number UPP site municipality municipality UFA centroid (mi) 1 Front Royal Falls Church 53.17 2 Fredericksburg Falls Church 42.94 3 Charlottesville Charlottesville 0.78 4 Harrisonburg Charlottesville 35.61 5 Roanoke Roanoke 5.51 6 Roanoke Roanoke 5.4 7 Charlottesville Charlottesville 1.37 8 Charlottesville Charlottesville 3.44 9 Alexandria Falls Church 8.21 10 Charlottesville Charlottesville 0.82 11 Charlottesville Charlottesville 0.82 12 Charlottesville Charlottesville 1.02 13 Charlottesville Charlottesville 0.95 14 Roanoke Roanoke 10.63 15 Alexandria Falls Church 8.22 16 Charlottesville Charlottesville 1.02

Table 21. Summary of species composition variables for on-site inventories of Urban Participant Properties. Variable Description Source Abundance Number of trees On-site inventory data Proportional contribution Calculated from site inventory Relative Abundance (RA) towards total site stem data count Formula-derived estimate Calculated by Eco1 using Structural Value (SV) of tree replacement cost CTLA2 formula Proportional contribution Calculated from site inventory Relative Structural Value (RSV) towards total site SV data Calculated by Eco using tree Summed total leaf Leaf Area (LA) dimensions and allometric surface area relationships Proportional contribution Relative Leaf Area (RLA) Calculated from Eco output towards total site LA Importance Value (IV) Sum of RA and RLA Calculated from Eco output Proportional contribution Calculated from Eco output; Relative Importance Value (RIV) towards total site IV equal to IV/2 1i-Tree Eco urban forest modeling software 2Council of Tree and Landscape Appraisers

136

Table 22. Comparisons of inventoried and non-inventoried Urban Participant Properties. P-value represents significance of test statistic from a comparison of means, using Welch’s t-test for continuous data and chi-square tests for proportional data. Inventoried sites (n = 16) Non-inventoried sites (n = 12) min max median SD x̅ min max median SD x̅ p Parcel size1 (acres) 0.13 1.67 0.53 0.41 0.60 0.14 2.04 0.39 0.53 0.57 0.847 Parcel tree cover2 (%) 0.14 0.91 0.58 0.24 0.54 0.12 0.67 0.41 0.18 0.38 0.053 Years since EAB detection1 2.0 11.0 2.5 3.2 4.1 0.0 11.0 2.5 3.2 3.6 0.659 Years since home 0.17 6.80 4.44 2.17 3.20 0.17 7.15 4.44 2.21 4.11 0.683 construction1 Historical ash relative 12.00 146.00 42.00 37.61 59.25 19.00 95.00 63.50 22.17 64.67 0.289 abundance2 (%) 1Means compared with Welch’s t-test. 2Means compared with chi-square test.

Table 23. Summary of ash species composition, sampling methods, and relative error for Urban Forest Assessments, by municipality. Roanoke Charlottesville Falls Church Year of data collection 2010 2011 2011 Total number of 0.1 acre plots 171 74 38 Stratified by land use yes yes no Total residential 0.1 acre plots 83 28 -- Ash Relative Abundance2 1.17%2 3.85%2 0%3 Relative S.E of ash abundance estimate1 75.00% 64.50% -- Ash Relative Structural Value 0.55%2 10.43%2 0%3 Relative S.E. of ash structural value estimate1 80.10% 95.70% -- Number of paired UHP properties 3 8 2 Mean distance to UHP properties (mi) 7.18 1.28 9.42 1Relative standard errors calculated as S.E./estimate*100 2Reported only for residential land use 3Reported for all land uses

137

Table 24. Tree benefits by species across 16 Urban Participant Properties. Top ten species, by Total Annual Benefits are listed. C Total Total C Gross C Runoff Pollution Energy emissions Annual Mean annual Number storage seq. avoided removal savings avoided Benefits benefits (per of trees (lbs) (lb/yr) (ft³/yr) (oz/yr) ($) (lb/yr) ($) tree, $) Platanus occidentalis 6 645.03 17.82 26.31 8.56 0.94 2.44 99.59 16.60 Fraxinus americana 59 16453.93 251.27 212.34 69.08 9.75 23.38 954.30 16.17 Quercus palustris 10 2528.53 54.70 36.57 11.89 0.96 1.16 161.65 16.17 Acer saccharum 8 750.62 21.94 27.76 9.04 1.39 2.99 115.23 14.40 Acer saccharinum 8 1582.22 19.36 27.73 9.02 1.24 2.84 109.96 13.75 Liriodendron tulipifera 14 2400.94 44.82 52.03 16.94 0.41 0.93 179.74 12.84 Juglans nigra 9 968.59 20.74 37.74 12.28 -0.61 -2.76 100.11 11.12 Quercus rubra 9 1165.40 23.95 20.07 6.53 1.19 2.86 95.45 10.61 Pinus strobus 17 383.83 12.39 27.40 8.93 3.93 11.79 154.28 9.08* Acer rubrum 32 2048.79 57.08 45.60 14.83 0.13 -1.99 169.48 5.30*** Bolded values show significant treatment contrast from the reference level of F. Americana *p<.05 ***p<.001

138

Table 25. Summary of site characteristics and species composition of inventoried Urban Participant Properties (n=16). Parcel size and tree cover were compiled from public databases; the remaining data were derived from data collected on-site. Parcel Number Ash Ash Parcel Parcel Total Site tree of trees Species Diversity Evenness Relative Relative legal inventoried Basal Number cover < 4" richness (H')1 (J')2 Abundance Basal Area acres acres Area (ft2) (%) DBH (%) (%) 1 0.42 0.18 91.2 21 17.6 9 2.06 0.94 9.5 10.2 2 0.38 0.38 32.1 28 19.0 13 2.12 0.83 7.1 47.4 3 0.40 0.40 48.6 32 57.0 10 1.81 0.79 21.9 47.7 4 0.69 0.69 57.5 47 42.3 15 2.38 0.88 14.9 29.6 5 0.52 0.52 27 18 11.5 5 0.84 0.52 5.6 27.0 6 0.70 0.70 13.9 13 18.3 10 2.25 0.98 15.4 34.4 7 0.71 0.31 57.9 15 21.2 10 2.18 0.95 13.3 27.8 8 1.04 0.56 79.2 27 49.6 12 2.37 0.96 11.1 29.6 9 0.69 0.50 77.6 32 49.7 17 2.72 0.96 3.1 2.8 10 0.14 0.22 23.5 7 19.7 6 1.75 0.98 14.3 46.7 11 0.55 0.55 27.7 9 35.0 9 2.20 1.00 11.1 26.3 12 0.30 0.30 51.4 8 35.5 6 1.73 0.97 25.0 59.7 13 0.14 0.14 69.1 4 20.2 3 1.04 0.95 50.0 62.4 14 1.67 2.05 67.3 42 122.1 11 1.99 0.83 38.1 68.1 15 1.16 0.82 88.7 43 62.1 14 2.26 0.86 23.3 24.3 16 0.31 0.31 57.8 19 33.4 11 2.20 0.92 5.3 6.6 1Species diversity measured by Shannon's diversity index (H') normally ranges from 1.5 to 3.5 2Species evenness measured by Pielou's J' ranges from 0 to 1

139

APPENDIX B Supplementary Tables for Chapter 4: Homeowner Engagement in Tree Preservation

List of Tables

Table 26. List of data sources employed in construction of general household survey sampling frame...... 141

Table 27. List of county and municipal data sources for land use zoning and property ownership, used in general household sampling...... 142

Table 28. Summary of survey items and measurement scales for homeowner surveys ...... 143

Table 29. Tests for nonresponse bias among program participant and general household sampling frames ...... 146

Table 30. Exploratory Factor Analysis for aggregated homeowner response data: Loadings of attitudinal statements on two factors...... 147

Table 31. List of predictors used in models of tree preservation intention and program participation ...... 148

Table 32. Program participants: Ranking of candidate models for the dependent variable Cost-share participation ...... 149

Table 33. Program participants: Summary of the top-ranked model Attitudes for the dependent variable Cost-share participation ...... 149

Table 34. General households: Ranking of candidate models for the dependent variable Cost-share participation ...... 150

Table 35. General households: Summary of the top-ranked model Attitudes for the dependent variable Cost-share participation ...... 150

140

Table 26. List of geographic criteria and data sources each were drawn from, employed in construction of the general household survey sampling frame. Variable Description Source Included Census Block Groups that TIGER shapefiles (U.S. Urban location are at least 75% urban, as measured Bureau of the Census 2017) by Census Urban Areas polygons. Counties stratified by EAB infestation: Established (pre-2015), VDOF records (Chamberlin EAB infestation strata Recent (2015 to present), and 2018) Undetected. Included residential land use only; Multiple county and Residential zoning mixed residential/commercial areas municipal tax parcel were excluded. databases (see Table 26)

Single-family Included parcels with only one VGIN Address Points dwellings associated address. Geodatabase (VGIN 2018)

Included tax parcels for which owner's Multiple county and Owner occupancy mailing address matched the parcel municipal tax parcel address. databases (see Table 26)

Attribute of VGIN Parcels Parcel size Minimum parcel area of 0.2 acres. Geodatabase (VGIN 2018)

Virginia Statewide Land Tree cover Minimum parcel tree cover of 25% Cover Database (VGIN 2018)

141

Table 27. List of county and municipal data sources for land use zoning and property ownership, used in general household sampling frame construction. Jurisdiction Data source URL City of Falls Church

City of Falls Church http://www.fallschurchva.gov/158/Maps Geospatial Services Arlington County

Arlington County https://gis.arlingtonva.us/GIS/gis_mappingcenter.asp GIS Mapping Center ArcGIS Open Data Hub -

City of Alexandria http://hub.arcgis.com/datasets/AlexGIS::parcels Alexandria Parcels

Roanoke County Roanoke County GIS Services https://www.roanokecountyva.gov/index.aspx?nid=76

City of Roanoke City of Roanoke GIS Services https://www.roanokeva.gov/518/Geographic-Information-Systems-GIS

City of Salem https://salemva.gov/Departments/Community- City of Salem

Engineering Department Development/Engineering ArcGIS Open Data Hub - http://hub.arcgis.com/datasets/HRPDC-GIS::hampton-roads-regional- City of Portsmouth

Hampton Roads Regional Parcels parcels ArcGIS Open Data Hub - http://hub.arcgis.com/datasets/HRPDC-GIS::hampton-roads-regional- City of Chesapeake

Hampton Roads Regional Parcels parcels

142

Table 28. Summary of survey items and measurement scales for homeowner surveys. Language differing between the two survey instruments is noted by ‘PP’ for items addressed only to participants and ‘GH’ for items addressed only to general households Survey items Variable class Measurement

PP: Likelihood of regularly treating at least one ash tree, for the foreseeable future Tree 5-point scale: 1= 'Extremely unlikely', 2= preservation 'Somewhat unlikely', 3= 'Neither likely nor unlikely', GH: Likelihood of regularly treating at least one intention 4= 'Somewhat likely', 5= 'Extremely likely' threatened landscape tree, for the foreseeable future PP: Likelihood of re-applying to the EABTP Cost-share GH: Likelihood of applying to an EABTP-like program Participation for other threatened landscape trees 6-point scale: 1= 'Under 24', 2= '26 to 34', 3= '35 to What is your age? Personal characteristics 44', 4= '45 to 54', 5= '55 to 64', 6= '65 or older' How many people are in your household? & activities Numerical entry How many in your household are under the age of 18? Numerical entry Which gender do you most identify with? Categorical: 'Male', 'Female', 'Other, please specify' Categorical: 'White', 'Black or African American', 'American Indian or Alaska Native', 'Asian', 'Native Which of the following best describe your racial or ethnic Hawaiian or Pacific Islander', Spanish, Latino or identity? Hispanic', 'Middle Eastern or North African', 'Other, please specify' Is English your first language? Categorical: 0= 'No', 1='Yes' 7-point scale: 1= 'Less than high school', 2= 'High What is the highest level of education you've had the school graduate', 3= 'Some college', 4= '2-year opportunity to complete? degree', 5= '4-year degree', 6= 'Graduate/Professional degree', 7= 'Doctorate'

143

8-point scale: 1= '$15,000 or less', 2= '$15,001 to $25,000', 3= '$25,001 to $35,000', 4= '$35,001 to What was your household's annual income in 2017 from $50,000', 5= '50,001 to $75,000', 6= '$75,001 to all earners and all sources, before taxes? $100,000', 7= '$100,000 to $150,000', 8= 'Greater than $150,000'

Property ownership (GH only) Categorical: 0= 'Renter', 1='Owner' 7-point scale: 1= '< 1 year', 2= '1 to 5 years', 3= '6 to How long have you lived at this property? 10 years', 4= '11 to 15 years', 5= '16 to 20 years', 6= '21 to 25 years', 7= '> 25 years' 7-point scale: 1= '0', 2= '$1 - $50', 3= '$51 to $250', In a typical year, how much is spent on tree maintenance? 4= '251 to $750, 5= '$751 to $1500', 6= '$1501 to $2500', 7= 'Greater than $2500' Before 2018, had you ever paid for insecticidal treatment Categorical: 0= 'No', 0= 'Not sure,' 1='Yes' of trees? Are you familiar with the emerald ash borer? (GH only) Categorical: 0= 'No', 1='Yes' Are there any ash trees on your property? (GH only) Categorical: 0= 'No', 0= 'Not sure,' 1='Yes' Before 2018, had you ever paid for treatment of ash trees? Categorical: 0= 'No', 0= 'Not sure,' 1='Yes' Have you heard of VDOF's Emerald Ash Borer Treatment Categorical: 0= 'No', 1='Yes' Program? Categorical: 'From a VDOF employee', 'Through a If you know about the EABTP, please describe how you VDOF publication', 'From the VDOF website', heard of it. 'From a friend or associate' Categorical: 'Trees are unlikely to survive', 'EAB is If you knew of the program, were eligible, yet decided not not yet in this region', 'Treatment is too costly', to apply, what best describes your reason? 'Didn't have enough information' 8-point scale: 1= 'Before 1950s', 2= '1950s', 3= Property When was the house built? '1960s', 4= '1970s', 5= '1980s', 6= '1990s', 7= characteristics '2000s', 8= '2010s'

144

About how many trees are on the property in 4-point scale: 1= '0 trees', 2= '1 to 5', 3= '6 to 10', 4= unmaintained areas? '> 10' About how many trees are on the property in maintained 4-point scale: 1= '0 trees', 2= '1 to 5', 3= '6 to 10', 4= areas? '> 10' 5-point scale: 1= 'Strongly disagree', 2= 'Somewhat How strongly would you agree or disagree with these Attitudes & disagree', 3= 'Neither agree nor disagree', 4= statements about urban trees: motivations 'Somewhat agree', 5= 'Strongly agree' Trees are an important part of the character of a neighborhood. Shade trees add value to a residential property. Overall, shade tree benefits outweigh hazards or nuisances The condition of trees on my property matters more than the turf I'd rather spend money to preserve than to remove an old tree. 5-point scale: 1= 'Not at all important', 2= 'Slightly How would you rate the importance of preserving ash important, 3= 'Somewhat important', 4= 'Very trees (GH: landscape trees) in the state of Virginia? important', 5= 'Extremely important' Rank the importance of the following reasons for preserving ash tree (GH: landscape trees), from most Ranked-choice item important (1) to least important (6) Shade Wildlife habitat Contribution to property value Tree species might become rare or endangered Trees are an attractive part of landscaping Other, please specify

145

Table 29. Tests for nonresponse bias among program participant and general household sampling frames. Test statistics display Welch's t-test and Pearson's χ2. Program participants (n=90) General households (n=1488) Non- Non- Respondents Test Respondents Test respondents p respondents p (n=53) statistic (n=333) statistic (n=39) (n=1155)

Parcel size (acres) 43.8 45.4 t=1.752 0.953 0.469 0.401 t= 1.75 0.0806

Parcel tree cover 0.589 0.526 χ2=0.181 0.671 0.45 0.452 χ2<.001 0.986 (%)

Mean of Census Block Group 44.6 44.9 t= -0.201 0.841 42.6 41.9 t=-1.4323 0.153 median age1 Mean of Census Block Group 78,706 72,693 t=1.13 0.26 116,197 104,966 t=2.5686 0.0105 median household income1 ($) Census Block Group mean years 15.3 14.8 t=1.72 0.0891 15.8 15.5 t=4.12 <.001 of education1

Census Block Group minority 10.2 13.1 χ2=0.0185 0.9 17.6 24.0 χ2= 5.48 0.0192 proportion1 (%)

Bold text indicates a significant result at the Bonferroni-corrected α < .008 significance level. 1Values derived from U.S. Bureau of the Census (2017) statistics for the Census Block Groups in which residential parcels were located

146

Table 30. Exploratory Factor Analysis for aggregated homeowner response data: Loadings of attitudinal statements on two factors. Bolded values indicate variable grouping for calculating factor indices. Factor 1 Factor 2 Attitudinal statements Tree curb appeal Tree affinity Trees are an important part of the character of a 0.849 neighborhood

Shade trees add value to a residential property 0.660 0.477

Overall, benefits provided by shade trees outweigh the 0.386 0.622 hazards and nuisances they can create. The condition of trees on my property is more important 0.717 than the condition of the turf. I'd rather spend money to preserve a mature tree than to 0.669 remove and replace it. Cronbach's alpha 0.790 0.670

147

Table 31. List of predictors used in models of tree preservation intention and program participation. Variables used only for program participant (PP) data and general households (GH) are noted. Predictors Predictor class Years living at property Personal characteristics Annual tree maintenance budget Prior treatment of landscape trees Familiarity with EAB Age Educational attainment Annual household income Age of house Property characteristics Number of trees in maintained areas Number of trees in unmaintained areas Parcel area Parcel tree cover Years since EAB detection (PP) EAB infestation strata - Undetected (GH) EAB infestation strata - Recent (GH) EAB infestation strata - Established (GH) Urban tree attitudes Factor A - Tree curb appeal Attitudes and motivations Urban tree attitudes Factor B - Tree affinity Importance of preserving landscape trees Tree preservation motivation - 2 clusters (PP) Tree preservation motivation - Cluster 1 (GH) Tree preservation motivation - Cluster 2 (GH) Tree preservation motivation - Cluster 3 (GH) Tree preservation motivation - Cluster 4 (GH) Tree preservation motivation - Cluster 5 (GH)

148

Table 32. Program participants: Ranking of candidate linear models for the dependent variable Cost-share participation. Number of respondents varied between models from 33 to 53. 1 2 2 3 Model K Adj. R RMSE ΔAICc wi Attitudes 4 0.223 0.717 0 0.996 Attitudes + Property 11 0.263 0.699 11.017 0.004 Property characteristics 7 -0.016 0.821 19.794 0 Personal characteristics 6 -0.073 0.843 28.041 0 Attitudes + Personal 10 0.037 0.799 33.774 0 Personal + Property 13 -0.292 0.925 53.394 0 Attitudes + Personal + Property 17 -0.19 0.888 72.359 0 Bold text indicates top-ranked model. 1Number of model parameters. 2Second-order AIC, a small sample-corrected version of AIC, a measure of model fit which minimizes information loss. 3Akaike’s weight, a proportional measure of relative likelihood.

Table 33. Program participants: Summary of the top-ranking linear model Attitudes for the dependent variable Cost-share participation (n=53) Independent variables 훽 coeff. p Urban tree attitudes Factor A - Tree curb appeal -0.167 0.405 Urban tree attitudes Factor B - Tree affinity 0.059 0.784 Importance of preserving ash trees 0.513 <.001 Tree preservation motivation - Cluster 1 -- -- Tree preservation motivation - Cluster 2 0.174 0.163 Adjusted 푅2 0.223 F 4.735 0.003 Bolded items indicate significance at the α=.05 level.

149

Table 34. General households: Ranking of candidate linear models for the dependent variable Cost-share participation. Number of respondents varied between models from 239 to 310. 1 2 2 3 Model K Adj. R RMSE ΔAICc wi Attitudes 7 0.456 0.853 0 0.936 Attitudes + Property 14 0.464 0.847 5.363 0.064 Attitudes + Personal 14 0.471 0.841 27.98 0 Attitudes + Personal + Property 21 0.469 0.843 37.485 0 Personal characteristics 7 0.065 1.118 147.968 0 Personal + Property 14 0.057 1.123 157.857 0 Property characteristics 7 0.056 1.123 167.618 0 Bolded text indicates top-ranked model. 1Number of model parameters. 2Second-order AIC, a small sample-corrected version of AIC, a measure of model fit which minimizes information loss. 3Akaike’s weight, a proportional measure of relative likelihood.

Table 35. General households: Summary of top-ranking linear model Attitudes for the dependent variable Cost-share participation (n=310) Independent variables 훽 coeff. p Urban tree attitudes Factor A - Tree curb appeal 0.11 0.044 Urban tree attitudes Factor B - Tree affinity 0.018 0.754 Importance of preserving landscape trees 0.609 <.001 Tree preservation motivation - Cluster 1 -- -- Tree preservation motivation - Cluster 2 -0.012 0.784 Tree preservation motivation - Cluster 3 0.06 0.207 Tree preservation motivation - Cluster 4 0.081 0.078 Tree preservation motivation - Cluster 5 -0.026 0.544 Adjusted 푅2 0.456 F 37.584 <.001 Bolded items indicate significance at the α=.05 level.

150

APPENDIX C Supplementary Tables for Chapter 5: Perceptions of Cost-Share Participation Among Forest Practitioners

List of Tables

Table 37. Summary of survey items and measurement scales for practitioner surveys ...... 152

Table 38. Results from tests for nonresponse bias among respondents to foresters’ survey ...... 155

Table 39. Tests for nonresponse bias among respondents to arborists' surveys ...... 155

Table 40. Exploratory Factor Analysis for aggregate practitioner response data: Loadings of attitudinal statements on two factors...... 156

Table 41. Aggregated forester and arborist response data: Cluster analysis of ranked statements of motivations for ash preservation ...... 156

Table 42. Arborist response data: Cluster analysis of ranked EAB management recommendations ...... 156

Table 43. List of predictors used in models of program participation for arborist response data, organized by predictor class ...... 157

151

Table 37. Summary of survey items and measurement scales for practitioner surveys. Language and items differing between the two surveys are noted by ‘FOR’ for items addressed only to foresters, and by ‘ARB’ for items addressed only to arborists. Survey items Variable class Measurement 5-point scale: 1= 'Not at all interested', 2= 'Slightly How would you rate your level of interest in participating in Cost-share interested, 3= 'Somewhat interested', 4= 'Very EABTP in future years? Participation interested', 5= 'Extremely interested'

Please describe your reasons for interest, or lack of interest, Written response in participating in this program. Personal What is your age? 6-point scale: 1= 'Under 24', 2= '26 to 34', 3= '35 to characteristics 44', 4= '45 to 54', 5= '55 to 64', 6= '65 or older'

Which gender do you most identify with? Categorical: 'Male', 'Female', 'Other, please specify' 7-point scale: 1= 'Less than high school', 2= 'High What is the highest level of education you've had the school graduate', 3= 'Some college', 4= '2-year opportunity to complete? degree', 5= '4-year degree', 6= 'Graduate/Professional degree', 7= 'Doctorate'

How many years of experience do you have working in Professional 8-point scale: 1='Less than 1', 2='1 to 5', 3='6 to 10', your field? characteristics 4='11 to 15', 5='16 to 20', 6='21 to 30', 7='31 to 40', 6='41+' Categorical: 0= 'No', 1= 'Previously', 1= 'Yes, Are you an SAF Certified Forester? currently' Categorical: 0= 'No', 1= 'Previously', 1= 'Yes, Are you an ISA Certified Arborist? currently' Categorical: 'Senior Area Forester', 'Area Forester What is your position at VDOF? (FOR) Specialist', 'Area Forester', 'Technician', 'Other' Categorical: 'Utility arboriculture', 'Commercial What is your area of practice? (ARB) arb.', 'Consulting or training', 'Urban forestry', 'Other' Is your business or agency located in Virginia? (ARB) Categorical: 0= 'No', 1= 'Yes'

152

If outside of Virginia, do you have clients in Virginia? Categorical: 0= 'No', 1= 'Yes' (ARB) In which county or independent city do you primarily work? Categorical: list of 135 Virginia jurisdictions What percentage of the property owners you work with own Percentage: 0 -100% > 5 acres? What percentage of your working hours are spent on forest Percentage: 0 -100% health projects?

Prior to 2018, how frequently were you dealing with EAB- 5-point scale: 1= 'Very rarely', 2= 'Rarely', 3= infested ash trees in your work? 'Occasionally', 4= 'Frequently', 5= 'Very frequently' Did you conduct any site vists as part of the 2018 EABTP? Numeric entry (FOR) Are you aware of VDOF's EABTP? (ARB) Categorical: 0= 'No', 1= 'Yes' Categorical: 'From a VDOF employee', 'VDOF How did you hear about the program? (ARB) publication', 'VDOF website', 'A friend or associate' How many bids for ash tree treatment, if any, did you Numeric entry submit as part of EABTP? (ARB) 5-point scale: 1= 'Strongly disagree', 2= 'Somewhat How strongly would you agree or disagree with these Attitudes & disagree', 3= 'Neither agree nor disagree', 4= statements about urban trees: motivations 'Somewhat agree', 5= 'Strongly agree' Trees are an interested part of the character of a neighborhood.

Shade trees add value to a residential property. Overall, shade tree benefits outweigh hazards or nuisances The condition of trees on my property matters more than the turf I'd rather spend money to preserve than to remove a declining tree.

153

Rank the importance of the following reasons for preserving ash trees in urban forests, from most important (1) to least Ranked-choice item important (6).

Shade Tree species might become rare

Wildlife habitat Attractive part of landscaping

Other, please specify Property value contribution Rank the relative level of threat from the following insect species to the forests in your region, from greatest level (1), Ranked-choice item to lowest level (7).

Hemlock woolly adelgid Southern pine beetle

Asian longhorned beetle Emerald ash borer

Gypsy moth Walnut twig beetle

Spotted lanternfly

Rank the following species according their suitability as residential shade trees, from most suitable (1) to least Ranked-choice item suitable (9). Ignore the effects of pests.

Sugar maple Green ash

White oak American elm

Black cherry Yellow-poplar

Black walnut Eastern white pine Rank the following EAB management strategies by how often you have recommended them, from most often (1) to Ranked-choice item least often (5). (ARB)

Tree removal Wait and see

Imadicloprid soil application Other, please specify

Emamectin trunk injection How do you expect sales of ash tree treatments would be 5-point scale: 1= 'Large decrease in sales', 2= affected if homeowners in your service area were widely 'Small decrease', 3= 'No effect', 4= Small increase', aware of EABTP funding? 5= 'Large increase in sales'

154

Table 38. Results from tests for nonresponse bias among respondents to foresters’ survey: chi- squared tests of independence between characteristics of survey respondents and a reference group of VDOF county foresters with equivalent positions. Reference group data are drawn from online VDOF records. Table displays observed frequencies by forester group for gender, job title, and region. VDOF Survey Pearson's Total p records respondents χ2 Gender Male 55 14 69 Female 9 4 13 0.701 0.402 Total 64 18 82 Job title Area Forester 43 12 55 Sr. Area Forester 21 6 27 0.002 0.967 Total 64 18 82 Region Western 17 4 21 Central 25 5 30 2.101 0.350 Eastern 22 10 32 Total 64 19 83 Bolded values display row and column totals.

Table 39. Tests for nonresponse bias among respondents to arborists' surveys: chi-squared tests of independence between characteristics of survey respondents and a reference group of MAC-ISA arborists in Virginia. Reference group data are drawn from records of MAC-ISA members within Virginia. Table displays observed frequencies by arborist group for area of practice and EAB infestation strata. MAC-ISA Survey Pearson's Cramer's Total p records respondents χ2 V Residential/ 131 (1.1) 46 (-1.1) 177 Commercial Municipal/ Area 96 (-1.6) 50 (1.6) 146 of Government 38.036 <.001 0.281 prac- Utility/Vegetation 42 (-0.2) 18 (0.2) 60 tice Research/Training 17 (-4.3) 24 (4.3) 41 Other 55 (4.5) 2 (-4.5) 57 Total 341 140 481 Established 277 (-1.0) 66 (1.0) 343 EAB Recent 196 (-1.2) 49 (1.2) 245 7.484 0.024 0.102 strata Undetected 122 (2.7) 13 (-2.7) 135 Total 595 128 723 Values in parenthesis represent standardized residuals. Bolded values display row and column totals.

155

Table 40. Exploratory Factor Analysis for aggregated practitioner response data: Loadings of attitudinal statements on two factors. Bolded values indicate variable grouping for calculating factor indices. Factor 1 Factor 2 Attitudinal statements Tree curb appeal Tree affinity Trees are an important part of the character of a 0.899 0.308 neighborhood Shade trees add value to a residential property 0.871 0.400 Overall, benefits provided by shade trees outweigh 0.329 0.507 the hazards and nuisances they can create. The condition of trees on my property is more 0.565 important than the condition of the turf. I'd rather spend money to preserve a mature tree than 0.677 to remove and replace it. Cronbach's alpha 0.670 0.950 Bolded values indicate variable grouping for calculating factor indices.

Table 41. Aggregated forester and arborist response data: Cluster analysis of ranked statements of motivations for ash preservation. Cluster 1 Cluster 2 Cluster 3 Cluster 4 Number of respondents 29 23 85 25 Proportion of survey respondents 0.180 0.142 0.524 0.154 Probability of membership 0.839 0.880 0.817 0.912 Modal order of ranked beliefs 1,4,2,5,3,6 1,3,5,2,4,6 4,1,2,3,5,6 4,5,1,2,3,6

Motivations were given as: 1) Ash trees provide shade, 2) Ash trees provide wildlife habitat, 3) Ash trees can increase property values, 4) Ash trees might become rare or endangered, 5) Ash trees are an attractive part of landscaping, 6) Other, please specify.

Table 42. Arborist response data: Cluster analysis of ranked EAB management recommendations Cluster 1 Cluster 2 Number of respondents 58 86 Proportion of survey respondents 0.402 0.598 Probability of membership 0.917 0.850 Modal order of ranked recommendations 1,4,2,3,5 4,1,2,3,5 Recommendations were given as: 1) tree removal, 2) imadicloprid trunk or soil application, 3) emamectin trunk injection, 4) wait and see, 5) Other, please specify.

156

Table 43. List of predictors used in models of program participation for arborist response data organized by predictor class. Predictors Predictor class Age Personal characteristics Gender Educational attainment Years of professional experience Professional characteristics Percentage of clientele with > 5 acres Percentage of forest health work hours Frequency of prior EAB-related work EAB management recommendation (2 clusters) EAB infestation strata - Undetected (GH) EAB infestation strata - Recent (GH) EAB infestation strata - Established (GH) Urban tree attitudes Factor A - Tree curb appeal Attitudes and motivations Urban tree attitudes Factor B - Tree affinity Expected effect of EABTP on sales Tree preservation motivation – 2 clusters (PP) Tree preservation motivation - Cluster 1 (GH) Tree preservation motivation - Cluster 2 (GH) Tree preservation motivation - Cluster 3 (GH) Tree preservation motivation - Cluster 4 (GH) Tree preservation motivation - Cluster 5 (GH)

157

APPENDIX D: 2018 EABTP Application Form

158

159

APPENDIX E: Tree Inventory Data Collection Form for Urban Participant Properties

160

APPENDIX F: Survey Recruitment Materials

Introductory Email for EABTP Participant Survey

161

Introductory Letter for General Household Survey

162

Introductory Email for VDOF Forester Survey

163

Introductory Email for MAC-ISA Arborist Survey

164

APPENDIX G: Survey Instruments

List of Surveys

EABTP Participant Survey ...... 166

General Household Survey ...... 173

VDOF Forester Survey ...... 181

MAC-ISA Arborist Survey ...... 187

165

EABTP Participant Survey

166

EABTP Participant Survey

167

EABTP Participant Survey

168

EABTP Participant Survey

169

EABTP Participant Survey

170

EABTP Participant Survey

171

EABTP Participant Survey

172

General Household Survey

173

General Household Survey

174

General Household Survey

175

General Household Survey

176

General Household Survey

177

General Household Survey

178

General Household Survey

179

General Household Survey

180

VDOF Forester Survey

181

VDOF Forester Survey

182

VDOF Forester Survey

183

VDOF Forester Survey

184

VDOF Forester Survey

185

VDOF Forester Survey

186

MAC-ISA Arborist Survey

187

MAC-ISA Arborist Survey

188

MAC-ISA Arborist Survey

189

MAC-ISA Arborist Survey

190

MAC-ISA Arborist Survey

191

MAC-ISA Arborist Survey

192

APPENDIX H: Western IRB Exemption Letter

193

194